Development, implementation, and usability evaluation of an mHealth-based clinical decision-support system for strengthening antimicrobial stewardship among physicians in Pakistan: a mixed-methods study Authors: Nosheen Nasir, Saadia Sattar, Asma Riaz, Abdul Muqeet, Noreen Zia, Hina Awat, Zara Nadeem, Saadbin Zafar, Beenish Khalid, Warda Imran Muallim, Salik Sadruddin Merani, Fariha Ahmed, Warda Ghias, Hassan Farooqui, Saleem Sayani, Ayeesha Kamran Kamal Published: 2026 ABSTRACT AbstractObjective:Antibiotic resistance is a major global health threat, driven by inappropriate antimicrobial prescribing. Mobile health (mHealth)–based clinical decision-support systems (CDSS) may strengthen antibiotic stewardship, but evidence from resource-limited settings remains limited. This study reports the development and implementation of ComBact, an mHealth-based CDSS for seven infectious disease syndromes, examines physicians’ perceptions of its clinical utility, and evaluates system-level usability metrics at a private tertiary-care hospital in Pakistan.Methods:A convergent parallel mixed-methods design was used. Semi-structured focus group discussions (FGDs) were conducted with internal medicine physicians who had access to ComBact for one year. Transcripts were analyzed using Braun and Clarke’s thematic framework. Quantitative analysis included system-generated data on user registration, uptake, and detailed usage metrics. Findings from both components were integrated using a joint display.Results:61 clinicians registered for ComBact, all accessing the dashboard at least once. The app recorded 134 sessions, with a mean of 1.23 pathway revisits per session and a mean workflow duration of 1 minute 25 seconds; 32.7% of users completed a ComBact plan. Mobile phones accounted for 57.4% of access and desktops for 42.6%. Two FGDs were conducted with fifteen physicians (median age 31 years (IQR: 27–34 yr). Five themes emerged: (1) clinical applicability, (2) user experience, (3) need of the app, (4) content limitations, and (5) recommendations.Conclusions:ComBact was perceived as a user-friendly tool with meaningful potential to support antimicrobial stewardship in LMICs, though significant content and functional enhancements are needed. Introduction Antibiotic resistance is a major public health concern worldwide, mainly driven by inappropriate antibiotic prescription and use in healthcare facilities. 1,2 A global study found a 16% increase in antibiotic consumption between 2016 and 2023, 3 potentially leading to increased healthcare costs, greater resistance to multidrug-resistant bacteria, and poorer patient outcomes. Low-middle-income countries (LMICs) are particularly prone to antibiotic resistance due to large population and infectious disease burden, ease of access to antibiotics, 4 scant antibiotic stewardship programs, and limited access to proper diagnostics and specialists. 5 A 2020 meta-analysis estimated that 52% of outpatient primary care visits in LMICs resulted in an antibiotic prescription, with a high proportion being inappropriate. 6 Competence in diagnosing infections and using antimicrobials appropriately cannot improve when access to training opportunities is limited. 7 As technology evolves in healthcare and access to mobile phones by patients and physicians increases, mobile health (mHealth) becomes a promising option for augmenting antibiotic stewardship efforts. 8,9 Use of clinical decision-support systems (CDSS) has been shown to reduce antibiotic prescribing by 34%, 10 consistent improvements in prescribing appropriateness, fewer errors, 11 and a significant increase (7%–74%) in adherence to guidelines. 12,13 However, to comprehensively analyze the influence of antimicrobial stewardship (AMS) apps on antimicrobial prescribing, further research is required. 14 There are limited studies on the impact of mHealth technology on antibiotic stewardship efforts in resource-constrained settings.15,16 This study aimed to develop and implement ComBact, an mHealth-based CDSS, and examine physicians’ perceptions of its usability, clinical utility and system-level usability metrics at a tertiary-care hospital in Pakistan. Methods Study setting and participants This study was conducted at Aga Khan University Hospital (AKUH), a tertiary-care academic medical center located in Karachi, Pakistan between January 2023 and June 2025. All physician prescribers (both in patient and outpatient) from the Section of Internal Medicine in Department of Medicine were invited to participate in this study. Nurse practitioners and pharmacists were excluded from this study. Study design This study employed a convergent parallel mixed-methods design. Qualitative component comprised of focus group discussions (FGDs) to capture in-depth user experiences, perceptions of usability of the mobile application, and its influence on their antibiotic prescribing practices. Quantitative analysis to characterize patterns of engagement comprised of application usage data, including frequency of access, feature utilization, and session duration. Following independent analysis, results from both study components were integrated. Clinical algorithm-based CDSS development The CDSS was developed by translating structured, evidence-based clinical algorithms into a digital platform tailored to the local context. Disease-specific algorithms for certain common infectious syndromes including skin and soft tissue infections, acute watery diarrhea (in immunocompetent), upper respiratory tract infections, urinary tract infection, acute undifferentiated fever (uncomplicated) (Typhoid fever), and complicated and uncomplicated Malaria, were incorporated, integrating clinical features with diagnostics (eg, C-reactive protein and cultures) to guide antibiotic decisions. These were aligned with institutional guidelines and refined using local antibiogram and antimicrobial resistance data. A multidisciplinary team, including infectious diseases specialists, internists, an AMS committee representative, and health informatics professionals, contributed to algorithm development, ensuring clinical validity, contextual relevance, and usability. The selected conditions were prioritized due to their high burden, frequent antibiotic misuse, and diagnostic uncertainty. Development of mHealth application “ComBact” The ComBact mHealth application was developed using a structured Software Development Life Cycle, including planning, design, development, testing, deployment, and maintenance (Figure 1). Requirements were defined through stakeholder consultations and workflow mapping. The interface replicated real-world inpatient and outpatient workflows with emphasis on usability and user experience and was built as a progressive web application using Angular and hosted on Google Firebase. The system used Google Firestore and BigQuery for reliable performance and data management. Core features included secure authentication, clinical assessment, treatment recommendation modules, and an AI-enabled server providing syndrome-based recommendations (Figure 2). Iterative testing ensured usability, reliability, and compliance before deployment across web and mobile platforms (Supplementary File 1). ComBact application’s user interface. Sequence diagram for ComBact. Implementation of mobile application Internal medicine physicians were informed through emails and meetings, and provided free access to the application via web-based deployment. The application supported Android, iOS, PC, and web platforms without requiring app store installation. Study team members provided onboarding and technical assistance. Participants accessed the intervention for 6–12 months. Data collection Data collection comprised two complementary components: FGDs: After at least 6 months to 1 year of application use, participants who met the eligibility criteria that is, those who had engaged in using the ComBact mobile application at least 3 times during the period, were invited to participate in FGDs to explore collective perceptions of the application’s usability, relevance to clinical workflow, perceived benefits and limitations, and suggestions for enhancement. Focus group guide was developed based on literature review and refined through pilot testing with clinicians (Supplementary File 1). These clinicians were subsequently removed from analysis. Discussions were moderated by experienced research faculty and conducted in English or Urdu according to participant preference, audio-recorded with informed consent. The interview was conducted at a learning venue in AKUH, lasting 30–40 minutes. The researcher used open-ended questions to elicit the physicians’ experiences. The audio recordings were then transcribed verbatim, translated into English and were subsequently reviewed against the original recordings to ensure accuracy and analyzed inductively using Braun and Clarke’s thematic framework. Both the audio recordings and the transcription were kept in a secure, password-protected folder with access limited to the research team. FGDs: After at least 6 months to 1 year of application use, participants who met the eligibility criteria that is, those who had engaged in using the ComBact mobile application at least 3 times during the period, were invited to participate in FGDs to explore collective perceptions of the application’s usability, relevance to clinical workflow, perceived benefits and limitations, and suggestions for enhancement. Focus group guide was developed based on literature review and refined through pilot testing with clinicians (Supplementary File 1). These clinicians were subsequently removed from analysis. Discussions were moderated by experienced research faculty and conducted in English or Urdu according to participant preference, audio-recorded with informed consent. The interview was conducted at a learning venue in AKUH, lasting 30–40 minutes. The researcher used open-ended questions to elicit the physicians’ experiences. The audio recordings were then transcribed verbatim, translated into English and were subsequently reviewed against the original recordings to ensure accuracy and analyzed inductively using Braun and Clarke’s thematic framework. Both the audio recordings and the transcription were kept in a secure, password-protected folder with access limited to the research team. Descriptive Usage Data: Engagement metrics were analyzed using principles from the AMUsED framework(17), which provides a structured approach for quantifying usage and exposure in digital interventions. Key metrics including session duration, total time in app, and sessions per user were calculated to assess user engagement. All reported “average” values represent the mean (arithmetic average), consistent with standard reporting practices for digital exposure metrics. Descriptive Usage Data: Engagement metrics were analyzed using principles from the AMUsED framework(17), which provides a structured approach for quantifying usage and exposure in digital interventions. Key metrics including session duration, total time in app, and sessions per user were calculated to assess user engagement. All reported “average” values represent the mean (arithmetic average), consistent with standard reporting practices for digital exposure metrics. Qualitative analysis All FGDs were transcribed verbatim and imported into NVivo (version 15.0, QSR International) for analysis following completion of data collection, which was continued until thematic saturation was achieved. Transcripts were coded independently by two members of the research team, including a trained qualitative researcher, to enhance analytical rigor. Coding was conducted iteratively, with initial transcripts used to develop a preliminary codebook that was refined as subsequent transcripts were analyzed. To ensure data familiarization, transcripts were read and re-read over a two-week period, and reflective field notes were triangulated with transcript data to support interpretation and coherence. The coded data were then reviewed by additional research team members (n = 3) to ensure consistency and reliability. An inductive thematic analysis was conducted using Braun and Clarke’s six-step framework. Emerging themes and sub-themes were iteratively developed, reviewed, and refined through discussions among the principal investigator and another co-investigator, with arbitration by a third investigator where needed to enhance credibility. The final thematic framework was agreed upon by consensus, and member checking was conducted with a subset of participating physicians further to ensure the trustworthiness and validity of the findings. Quantitative analysis Descriptive statistics were generated to characterize patterns of application engagement among study participants. Usage metrics extracted from the application backend included: number of logins per user, cumulative time spent on the app, frequency of feature usage, and temporal trends in access (eg, use by time of day). These metrics were summarized using means for continuous variables and counts with percentages for categorical usage categories. Following independent analysis of quantitative and qualitative data, findings were integrated using a joint display approach. Quantitative usage metrics were aligned with qualitative themes, illustrative quotations, and corresponding inferences to enable inferences across data sources. Ethical considerations Ethical approval was obtained from the Aga Khan University ethics review committee prior to the commencement of the study (ERC Reference number: 5487). All physicians were approached via email and electronic informed consent was obtained. Results Qualitative results A total of two FGDs were conducted for this study. The first focus group comprised internal medicine consultants, while the second included resident physicians. Altogether, 15 participants took part, with a median age of 31 years (IQR: 27–34 yr). This included four consultants (senior attending-level physicians), three instructors (early-career faculty), and eight residents (postgraduate trainees). Among the residents, there were two first-year, two second-year, one third-year, one fourth-year, and two chief residents who had completed their four-year internal medicine residency at the institute. The instructors had a minimum of 3.5 years of clinical experience, while the consultants had a median of 4 years of experience (IQR: 3–7.5 yr). We identified five interrelated themes that describe physicians’ experiences with the CDSS. Each theme reflects not only what participants reported, but how they interpreted the app’s role within their clinical workflow (Table 1). Table 1.Summarized qualitative findings; themes, sub-themes and physician’s illustrative quotations Table 1 long description. ThemesSub-themesPhysician’s views1. Outpatient-Focused Clinical Utility I. Comprehensive Management Plan II. Usability in Stable Patient Care Areas “I liked the straightforward options provided for malaria. First, we are provided with a standard treatment regimen. This worked well for me because it includes a review of the disease. Even if you are aware of the regimen, all the symptoms are revised as well. The app also provides treatment options for complicated malaria “(In the case of) cystitis, (the app provided the option of) nitrofurantoin, which I really liked.” When I selected urinary tract infection, I received different antibiotic choices. So, this was a positive for me.” “My personal opinion is that this (app) is more suited to an outpatient area or a general practice setting.” “I don’t think we can use this app for patients admitted in Special Care Unit (SCU) or Intensive Care Unit (ICU) as the complexity of illness increases (there).” 2. Workflow efficiency versus technical barriers I. User Friendly and Time Efficient II. Barriers to App’s User interface and Functionality “The app is user-friendly, with an easy-to-understand and easy-to-use pattern of options.” “The flow is easy. The options are presented clearly, followed by a ‘proceed next’ option that makes it easy to navigate.” “The pattern is good with respect to the patient’s symptoms and the required algorithm. It starts with history and follows a step-by-step Process.” “In this (app), all the steps are provided, including when and which antibiotics to prescribe, which benefits my patient. It is (a) very direct (app), in a short time- the decision is made.” “It is a fast app. The format for presenting options and moving forward is similar to artificial intelligence (AI) software.” “The single sign-in feature means my login is saved, so I don’t have to log in again every time.” “In many places, it seems that you reach a dead end. You click an option, and no further options are provided.” “After selecting an option, we are unable to proceed. The app returns to the first page and hangs up.” “The loading of the app takes time and we end up waiting during rounds.” “I like the flow, but it is not ideal. It should follow a sequence of symptoms, then labs, then diagnosis. The streamline is not as smooth as it should be.” “When we press ‘proceed next’, we often have to reload the page three or four times.” “When you click ‘the next’ option, the app prompts you to press ‘proceed’ but the high number of clicks required minimizes app usability.” 3. Need for integrated decision support I. Usage of a Single Source for Diagnosis and Management II. Resource Intensive Content Search III. Formulating Diagnosis and Recommending Treatment Plan “We currently spend time searching for guidelines online. We use Up-to-Date as authentic clinical information is present in one source, even though searching for specific antibiotic dosages and indications is time consuming.” “If you provide the content in the app, even if the interface is not user-friendly, we are still going to use it. We would also like the end-results—the clinical recommendations—to be patient- and physician-friendly.” “If I am getting the required clinical content, then I will use the app. We search the websites for the same reason. So, if we are getting something easily on mobile, we prefer to use it.” “With the (available) apps, we do have the guidelines but we must make the diagnosis ourselves. If we have an app where the patient’s symptoms are entered and it formulates the diagnosis, it becomes easier and useful. So, we will prefer that to help us see and manage the patient.” “If the app allows us to input a patient’s signs and symptoms and, in turn, recommends relevant antimicrobials, physicians would not mind using it even if the interface is somewhat difficult.” 4. Gaps in clinical scope and adaptability I. Lack of Coverage of Systemic Diseases II. Necessity of Inclusion of Clinical Predicaments III. Limitations in Options for Cultures and Variable Septic Marker Role “Most syndromes are not covered. Common diseases like meningitis, liver abscess, lower respiratory tract infections that we encounter are not included in this app.” “I was using the skin infections category. There were no options for cellulitis, which was what my patient had. The available options were for conditions like furuncles, carbuncles, eczema, and impetigo. I tried to select the most relatable options, but there was no guidance on antibiotics.” “For acute undifferentiated fever, the app focuses exclusively on enteric fever and does not provide options for other potential diagnoses. Similarly, for acute watery diarrhea, it does not offer treatment options for other causes, limiting its recommendations to antibiotics for enteric fever, such as ceftriaxone and meropenem.” “(While using this App) We encountered difficulty when we had patients with deranged creatine and liver enzymes as the app did not provide alternate antibiotic options for these patients.” “If someone is prescribing carbapenem to a patient and that patient is allergic to it, we don’t get any alternative option which is why this app is not feasible for us to use.” “We upgrade to broad-spectrum antibiotics when patients go into shock during admission, but this option is not available in the app. For patients with acute gastroenteritis, there are options for supportive treatment; however, if that patient develops septic shock, the app does not provide options for broad-spectrum antibiotics.” “One of my immunocompromised patients developed acute gastroenteritis, but the app provided treatment options for acute watery diarrhea in immunocompetent patients. The management for these scenarios is completely different and the app did not provide treatment options for immunocompromised individuals.” “The app does not provide options for cultures other than blood. For instance, even when selecting the complicated urinary tract infection option, it only suggests blood cultures and not urinary cultures.” “While other cultures may not be prioritized like blood cultures, following treatment guidelines requires that a positive urine culture not be ignored in a symptomatic patient.” “Under acute undifferentiated fever, the options provided are based solely on C-reactive protein (CRP) levels being within range or elevated. However, deciding whether to start antibiotics based only on this is not a good approach.” 5. Recommendations for enhancing functionality and relevance I. Patients’ Clinical Condition Led Management II. Referenced Clinical Content III. Features for Current Clinical Trends, Antibiotic Dosage Adjustment, and Resistance IV. Enhancement of App Processes “The app should be customized according to the variation of the patient symptoms and the clinical conditions.” “The app should specify antibiotics to prescribe to a patient who has failed an initial course.” ‘‘I think if citations are added, such as the guidelines from which the app derives its information, it would be helpful. That way, if one is unable to understand something or there is a problem with the flow, you can be directed to the authentic source to get verified information. I know that it might not be possible, but I think references should be there.” ‘‘It would be beneficial, specifically for antibiotics, if the app provided dose adjustments based on creatinine clearance.” “There should be a feature in this app for antibiotic resistance.” “I understand that this is a pilot model, but a (competitor) app has features for (clinical) calculations, current trends and adjusted antibiotic dosage. Additionally, the disease-related symptoms are explained comprehensively. Therefore, I recommend making this app similar to that model.” “On the starting page, there are eight to nine options for acute undifferentiated fever. When you click, it gives you symptoms based on your selection rather than going to a ‘next page’. This pattern should be consistent across the app; after selecting an option, the corresponding choices should be based on that selection.” “It would be more convenient if the corresponding page appeared immediately after selecting an option.” Summarized qualitative findings; themes, sub-themes and physician’s illustrative quotations Table 1 long description. I. Comprehensive Management Plan II. Usability in Stable Patient Care Areas “I liked the straightforward options provided for malaria. First, we are provided with a standard treatment regimen. This worked well for me because it includes a review of the disease. Even if you are aware of the regimen, all the symptoms are revised as well. The app also provides treatment options for complicated malaria “(In the case of) cystitis, (the app provided the option of) nitrofurantoin, which I really liked.” When I selected urinary tract infection, I received different antibiotic choices. So, this was a positive for me.” “My personal opinion is that this (app) is more suited to an outpatient area or a general practice setting.” “I don’t think we can use this app for patients admitted in Special Care Unit (SCU) or Intensive Care Unit (ICU) as the complexity of illness increases (there).” 2. Workflow efficiency versus technical barriers I. User Friendly and Time Efficient II. Barriers to App’s User interface and Functionality “The app is user-friendly, with an easy-to-understand and easy-to-use pattern of options.” “The flow is easy. The options are presented clearly, followed by a ‘proceed next’ option that makes it easy to navigate.” “The pattern is good with respect to the patient’s symptoms and the required algorithm. It starts with history and follows a step-by-step Process.” “In this (app), all the steps are provided, including when and which antibiotics to prescribe, which benefits my patient. It is (a) very direct (app), in a short time- the decision is made.” “It is a fast app. The format for presenting options and moving forward is similar to artificial intelligence (AI) software.” “The single sign-in feature means my login is saved, so I don’t have to log in again every time.” “In many places, it seems that you reach a dead end. You click an option, and no further options are provided.” “After selecting an option, we are unable to proceed. The app returns to the first page and hangs up.” “The loading of the app takes time and we end up waiting during rounds.” “I like the flow, but it is not ideal. It should follow a sequence of symptoms, then labs, then diagnosis. The streamline is not as smooth as it should be.” “When we press ‘proceed next’, we often have to reload the page three or four times.” “When you click ‘the next’ option, the app prompts you to press ‘proceed’ but the high number of clicks required minimizes app usability.” 3. Need for integrated decision support I. Usage of a Single Source for Diagnosis and Management II. Resource Intensive Content Search III. Formulating Diagnosis and Recommending Treatment Plan “We currently spend time searching for guidelines online. We use Up-to-Date as authentic clinical information is present in one source, even though searching for specific antibiotic dosages and indications is time consuming.” “If you provide the content in the app, even if the interface is not user-friendly, we are still going to use it. We would also like the end-results—the clinical recommendations—to be patient- and physician-friendly.” “If I am getting the required clinical content, then I will use the app. We search the websites for the same reason. So, if we are getting something easily on mobile, we prefer to use it.” “With the (available) apps, we do have the guidelines but we must make the diagnosis ourselves. If we have an app where the patient’s symptoms are entered and it formulates the diagnosis, it becomes easier and useful. So, we will prefer that to help us see and manage the patient.” “If the app allows us to input a patient’s signs and symptoms and, in turn, recommends relevant antimicrobials, physicians would not mind using it even if the interface is somewhat difficult.” 4. Gaps in clinical scope and adaptability I. Lack of Coverage of Systemic Diseases II. Necessity of Inclusion of Clinical Predicaments III. Limitations in Options for Cultures and Variable Septic Marker Role “Most syndromes are not covered. Common diseases like meningitis, liver abscess, lower respiratory tract infections that we encounter are not included in this app.” “I was using the skin infections category. There were no options for cellulitis, which was what my patient had. The available options were for conditions like furuncles, carbuncles, eczema, and impetigo. I tried to select the most relatable options, but there was no guidance on antibiotics.” “For acute undifferentiated fever, the app focuses exclusively on enteric fever and does not provide options for other potential diagnoses. Similarly, for acute watery diarrhea, it does not offer treatment options for other causes, limiting its recommendations to antibiotics for enteric fever, such as ceftriaxone and meropenem.” “(While using this App) We encountered difficulty when we had patients with deranged creatine and liver enzymes as the app did not provide alternate antibiotic options for these patients.” “If someone is prescribing carbapenem to a patient and that patient is allergic to it, we don’t get any alternative option which is why this app is not feasible for us to use.” “We upgrade to broad-spectrum antibiotics when patients go into shock during admission, but this option is not available in the app. For patients with acute gastroenteritis, there are options for supportive treatment; however, if that patient develops septic shock, the app does not provide options for broad-spectrum antibiotics.” “One of my immunocompromised patients developed acute gastroenteritis, but the app provided treatment options for acute watery diarrhea in immunocompetent patients. The management for these scenarios is completely different and the app did not provide treatment options for immunocompromised individuals.” “The app does not provide options for cultures other than blood. For instance, even when selecting the complicated urinary tract infection option, it only suggests blood cultures and not urinary cultures.” “While other cultures may not be prioritized like blood cultures, following treatment guidelines requires that a positive urine culture not be ignored in a symptomatic patient.” “Under acute undifferentiated fever, the options provided are based solely on C-reactive protein (CRP) levels being within range or elevated. However, deciding whether to start antibiotics based only on this is not a good approach.” 5. Recommendations for enhancing functionality and relevance I. Patients’ Clinical Condition Led Management II. Referenced Clinical Content III. Features for Current Clinical Trends, Antibiotic Dosage Adjustment, and Resistance IV. Enhancement of App Processes “The app should be customized according to the variation of the patient symptoms and the clinical conditions.” “The app should specify antibiotics to prescribe to a patient who has failed an initial course.” ‘‘I think if citations are added, such as the guidelines from which the app derives its information, it would be helpful. That way, if one is unable to understand something or there is a problem with the flow, you can be directed to the authentic source to get verified information. I know that it might not be possible, but I think references should be there.” ‘‘It would be beneficial, specifically for antibiotics, if the app provided dose adjustments based on creatinine clearance.” “There should be a feature in this app for antibiotic resistance.” “I understand that this is a pilot model, but a (competitor) app has features for (clinical) calculations, current trends and adjusted antibiotic dosage. Additionally, the disease-related symptoms are explained comprehensively. Therefore, I recommend making this app similar to that model.” “On the starting page, there are eight to nine options for acute undifferentiated fever. When you click, it gives you symptoms based on your selection rather than going to a ‘next page’. This pattern should be consistent across the app; after selecting an option, the corresponding choices should be based on that selection.” “It would be more convenient if the corresponding page appeared immediately after selecting an option.” Theme 1: outpatient-focused clinical utility Physicians consistently perceived the CDSS as a clinically useful tool for managing common, stable outpatient conditions. The app’s structured, algorithm-driven approach supported standardized assessment and treatment, reinforcing adherence to guideline-based care. Participants valued the ability to quickly access antibiotic recommendations tailored to specific clinical scenarios, which reduced cognitive burden during consultations. Physicians highlighted that while the app was well-suited for outpatient settings, its applicability diminished in high-acuity environments such as intensive care units. In these settings, clinical complexity, patient instability, and the need for individualized decision-making limited reliance on algorithm-based guidance. Theme 2: workflow efficiency versus technical barriers Physicians described the app’s interface as simple, easy to use, and logically sequenced. They reported that the app’s stepwise antibiotic selection contributed to more time-efficient clinical management. The ability to arrive at treatment decisions quickly was seen as a major advantage, particularly in time-constrained outpatient settings. Physicians shared several user interface challenges. Issues such as delays in loading, repeated clicks, navigation breakdowns, and system interruptions disrupted clinical workflow and reduced usability. These barriers not only affected efficiency but also undermined user confidence in the tool. Theme 3: need for integrated decision support Physicians emphasized the need for an all-purpose mobile application featuring updated guidelines and authenticated information on antibiotic indications and dosages. They expressed that even a subpar interface would be acceptable if the application provided necessary information to guide clinical decision-making. Physicians emphasized that current practice often involves consulting multiple resources, which is time-consuming and fragmented. In this context, the CDSS was valued for its potential to consolidate clinical guidance into one accessible tool. Importantly, participants expressed a preference for systems that go beyond static guideline presentation to actively support clinical reasoning. The ability to input patient-specific information and receive tailored diagnostic and therapeutic recommendations was seen as particularly beneficial. Theme 4: gaps in clinical scope and adaptability A key concern was the limited coverage of clinical conditions, particularly common or complex infections encountered in routine practice. This restricted the app’s usefulness and, in some cases, forced clinicians to rely on external resources. Additionally, the CDSS lacked adaptability to more complex patient scenarios, such as those involving comorbidities, immunocompromised states, organ dysfunction, or clinical deterioration. Participants also noted limitations in antibiotic selection logic, including insufficient alternatives for allergies, lack of escalation pathways, and over-reliance on single parameters (eg, CRP) in decision-making. These findings suggest that while algorithm-based systems can standardize care, their effectiveness depends on their ability to accommodate clinical variability and uncertainty. Theme 5: recommendations for enhancing functionality and relevance Based on their user experiences, physicians provided recommendations to enhance the app’s clinical content and functionality. Physicians emphasized the need to incorporate adjustments for chronic medical conditions like kidney and liver disease to provide dose-adjusted medications. They also stressed that the app’s core model should include features to counter antibiotic resistance, essential clinical calculations, and updated knowledge of diseases and current trends. The need for authentication via references was also relayed to ensure a comprehensive understanding of the disease management suggested. Quantitative results A total of 61 users registered for the app, and all of them accessed the app to view the dashboard and its information. There were 134 sessions conducted within the application, with each session being viewed an average of 1.23 times. This indicates that some users revisited the app process twice. User activation was 100% indicating frictionless onboarding and effective first-use dashboard. Session density was 2.2 sessions per user from 134 sessions and average session duration was 1 m 25 s (85s). Landing page dwell time was 18s and plan completion rate was 32.71% (Table 2). Table 2.Key performance indicators (KPI) summary Table 2 long description. MetricValueInterpretation User activation rate 100% (61/61)All registered users successfully accessed and viewed the initial content. Sessions per user ∼2.2 SessionsDemonstrates initial stickiness and re-engagement by the user base. Conversion rate (plan completion) 32.71%The percentage of sessions that completed the core ComBact pathway. Average session duration (process) 1 min 25 s (85s)Indicates the typical time required for users who engage with the antimicrobial pathway process. Active re-use rate >23% (14/61)At least 14 unique users returned after the first use, confirming value proposition. Platform distribution Mobile: 57.4%/Desktop: 42.6%Confirms a strong preference for mobile access. Key performance indicators (KPI) summary Table 2 long description. Discussion This mixed-methods study evaluated physicians’ experiences of the “ComBact” antibiotic stewardship app in a Pakistani tertiary hospital. Physicians found it intuitive, efficient, and clinically useful, but noted limitations in complex cases, technical performance, and critical-care integration. Analytics highlighted needs for workflow optimization, mobile testing, improved landing-page efficiency, and re-engagement strategies. The integration of quantitative usage metrics with qualitative themes and sub-themes provides a nuanced understanding of the app’s engagement, usability, and adoption patterns (Table 3). High user activation rates aligned with qualitative themes of app operability and user interface, suggesting initial usability and efficient clinical decision-making. This matched meaningful usage, reflected by session frequency and re-use rates. In contrast, prolonged page dwell aligned with workflow interruptions, navigation difficulties, and technical glitches, indicating usability friction. Low plan completion rates reflected limited disease coverage and insufficient comorbidity accommodation, constraining decision pathways within the app. Table 3.Joint display integrating quantitative indicators, qualitative themes, sub-themes, illustrative quotations and their corresponding inferences Table 3 long description. Quantitative indicatorsQualitative themes/sub-themes and illustrative quotationsInferences User activation rate: 100% Mean activation time: 1 minute 25 seconds Theme: App Operability and User Experience -User Friendly and Time Efficient “The app is user-friendly, with an easy-to-understand and easy-to-use pattern of options.” “The flow is easy. The options are presented clearly, followed by a ‘proceed next’ option that makes it easy to navigate.” The app’s easy-to-use interface and logical sequencing of options align with high activation rates, indicating strong initial usability and perceived relevance. Sessions per user: 2.2 Per User Re-use rate: 23% Theme: App Operability and User Experience -User Friendly and Time Efficient “The app is user-friendly, with an easy-to-understand and easy-to-use pattern of options.” ‘‘The flow is easy. The options are presented clearly, followed by a ‘proceed next’ option that makes it easy to navigate.” “In this (app), all the steps are provided, including when and which antibiotics to prescribe, which benefits my patient. It is (a) very direct (app), in a short time- the decision is made Themes: Clinical Applicability- Comprehensive Management Plan When I selected urinary tract infection, I received different antibiotic choices. So, this was a positive for me.” Qualitative findings indicate the app facilitated rapid clinical management and offered alternative antibiotic choices, which suggests selective but purposeful use aligned with point-of-care decision-making. Landing page dwell time: 18 s Theme: App Operability and User Experience- Barriers to Apps User Interface and Functionality “In many places, it seems that you reach a dead end. You click an option, and no further options are provided.” “After selecting an option, we are unable to proceed. The app returns to the first page and hangs up.” Theme: Content Limitations—Lack of Coverage of Systemic Diseases “Most syndromes are not covered. Common diseases like meningitis, liver abscess, lower respiratory tract infections that we encounter are not included in this app.” Prolonged dwell time, aligns with qualitative findings of technical glitches, dead-end navigation and limited disease coverage, suggesting that time spent on the landing page reflects workflow interruption and navigation difficulty rather than deliberate clinical reasoning. Plan completion rate-32.71% Theme: App Operability and User Experience- Barriers to Apps User Interface and Functionality “When we press ‘proceed next’, we often have to reload the page three or four times.” “The loading of the app takes time and we end up waiting during rounds.” Theme: Content Limitations—Lack of Coverage of Systemic Diseases “For acute undifferentiated fever, the app focuses exclusively on enteric fever and does not provide options for other potential diagnoses. Similarly, for acute watery diarrhea, it does not offer treatment options for other causes, limiting its recommendations to antibiotics for enteric fever, such as ceftriaxone and meropenem.” Theme: Content Limitations—Necessity of Inclusion of Clinical Predicaments “(While using this App) We encountered difficulty when we had patients with deranged creatine and liver enzymes as the app did not provide alternate antibiotic options for these patients.” The lower app completion rate aligns with qualitative reports of repeated app reloads and limited disease and comorbidity coverage, suggesting that both operability challenges and limited clinical scope constrained pathway completion. Joint display integrating quantitative indicators, qualitative themes, sub-themes, illustrative quotations and their corresponding inferences Table 3 long description. User activation rate: 100% Mean activation time: 1 minute 25 seconds Theme: App Operability and User Experience -User Friendly and Time Efficient “The app is user-friendly, with an easy-to-understand and easy-to-use pattern of options.” “The flow is easy. The options are presented clearly, followed by a ‘proceed next’ option that makes it easy to navigate.” Sessions per user: 2.2 Per User Re-use rate: 23% Theme: App Operability and User Experience -User Friendly and Time Efficient “The app is user-friendly, with an easy-to-understand and easy-to-use pattern of options.” ‘‘The flow is easy. The options are presented clearly, followed by a ‘proceed next’ option that makes it easy to navigate.” “In this (app), all the steps are provided, including when and which antibiotics to prescribe, which benefits my patient. It is (a) very direct (app), in a short time- the decision is made Themes: Clinical Applicability- Comprehensive Management Plan When I selected urinary tract infection, I received different antibiotic choices. So, this was a positive for me.” Theme: App Operability and User Experience- Barriers to Apps User Interface and Functionality “In many places, it seems that you reach a dead end. You click an option, and no further options are provided.” “After selecting an option, we are unable to proceed. The app returns to the first page and hangs up.” Theme: Content Limitations—Lack of Coverage of Systemic Diseases “Most syndromes are not covered. Common diseases like meningitis, liver abscess, lower respiratory tract infections that we encounter are not included in this app.” Theme: App Operability and User Experience- Barriers to Apps User Interface and Functionality “When we press ‘proceed next’, we often have to reload the page three or four times.” “The loading of the app takes time and we end up waiting during rounds.” Theme: Content Limitations—Lack of Coverage of Systemic Diseases “For acute undifferentiated fever, the app focuses exclusively on enteric fever and does not provide options for other potential diagnoses. Similarly, for acute watery diarrhea, it does not offer treatment options for other causes, limiting its recommendations to antibiotics for enteric fever, such as ceftriaxone and meropenem.” Theme: Content Limitations—Necessity of Inclusion of Clinical Predicaments “(While using this App) We encountered difficulty when we had patients with deranged creatine and liver enzymes as the app did not provide alternate antibiotic options for these patients.” A study from Italy explored 27 applications from 115 applications for appropriateness of content on antimicrobial prescribing to facilitate healthcare professionals and found that none of them fulfilled the criteria for having the required features which could be implemented highlighting the need for significant improvement in the development of mobile applications to assist prescribers, considering the local context. 18 The utility and impact of mobile CDSS to enhance antimicrobial stewardship efforts is largely unexplored in LMICs though there have been studies that have emphasized its potential in these settings. 19 Currently, there is no mobile application from Pakistan that focuses on antimicrobial prescribing and antimicrobial stewardship though Pakistan ranks third amongst LMICs in human antimicrobial consumption. 20 Hence, in contrast to those studies conducted in high-resource settings, our study extends the evidence by documenting implementation feasibility and user perceptions in a South Asian tertiary-care hospital. Our qualitative findings highlight the significance of end-user acceptability, workflow fit, and context as crucial enablers in the analysis. Participants in our study valued ComBact’s simplicity, logical flow, and time efficiency, echoing findings from Peiffer-Smadja et al., who found that usability and workflow alignment were key to adoption of CDSS in West African primary care settings. 19 Physicians also recognized the benefit of having an authentic, single-source reference for antibiotic guidance, especially since access to subscription-based platforms (eg, UpToDate, Sanford Guide) is limited. Nonetheless, users voiced concern about the insufficient content coverage, particularly for complex infections, immunocompromised patients, patients in critical care settings and illnesses. In a systematic assessment of stewardship applications, Helou et al concluded that most lacked adaptation for specific populations and pathogens. 14 These critiques are similar. Analysis of in-app user behavior revealed several key engagement patterns that help explain where friction occurs in the physician workflow and how the tool might be optimized. Notably, there was a substantial drop-off (67%), indicating a critical point where many users disengaged before completing guideline-based recommendations. This suggests potential usability bottlenecks, such as unclear navigation, excessive cognitive load, or complex input requirements. Targeted optimization strategies, including progress indicators, simplified input fields, and contextual guidance, may improve completion rates and streamline clinician experience. With mobile devices accounting for 57.4% of app access, mobile-specific usability emerged as another area requiring attention. The observed session density (2.2 sessions per user over one year) suggests modest but repeat engagement with the CDSS. While some users returned beyond initial use, the overall frequency remains low for a tool intended to support routine clinical decision-making. In health informatics, sessions per user serve as a proxy for sustained engagement and workflow integration; therefore, this finding likely reflects limited habitual adoption. Several factors may explain this pattern. First, the CDSS was designed for specific clinical scenarios rather than continuous use across all patient encounters, limiting expected use frequency. Second, qualitative findings suggest that technical barriers and limited clinical scope may have reduced repeated use. Third, digital tools must compete with established practices and time constraints in busy clinical environments, which may hinder consistent engagement. Low session frequency is not uncommon in mHealth interventions unless they are closely integrated into workflows or supported by reinforcement strategies. This highlights the need for enhanced integration, broader clinical applicability, and improved usability. Future iterations should incorporate workflow integration, performance optimization, and expanded clinical content to improve longitudinal engagement. Engagement metrics were summarized using mean values, which estimate overall user exposure. Although median values may provide additional insight into typical usage patterns, they were not available in the analytics data set. Nevertheless, mean values remain consistent with standard reporting approaches and are sufficient to characterize overall engagement trends in this study. To our knowledge, this is the first study in Pakistan to describe the co-development and qualitative evaluation of a stewardship mHealth app among practicing physicians. Using FGDs, app usage data, and physician feedback, this mixed-methods study provided a comprehensive assessment of user experience and feasibility, with findings from a real-world tertiary-care hospital in Pakistan offering valuable insights for LMIC settings where evidence remains limited. The study has several limitations, including potential volunteer bias due to self-selected participants who may have been more digitally inclined. Additionally, its conduct within a single tertiary-care institution may limit generalizability to other settings, and longer-term acceptability and sustained use were not evaluated. Although the clinical algorithms underwent iterative expert review and consensus-based validation before implementation, representing standard early-stage content validation approaches used in CDSS development, formal validation through clinical outcome studies was not performed. From a health informatics perspective, early evaluations of mHealth tools often rely on usage and usability metrics rather than clinical end points, and engagement measures primarily reflect interaction rather than effectiveness. Accordingly, the CDSS is intended as a decision-support adjunct, not a replacement for clinical judgment. Future work should focus on prospective validation of prescribing outcomes, safety, and patient-level impact. Overall, the ComBact mHealth app was perceived as a user-friendly and effective clinical decision support tool for antibiotic stewardship, demonstrating promise in LMIC settings and warranting further evaluation of its clinical and economic impact. Supporting information Nasir et al. supplementary material Nasir et al. supplementary material