A multicentric intervention based on nudging for the reduction of potentially unnecessary laboratory tests: insights from a Swiss hospital network Authors: Angela Greco, Maria Luisa Garo, Martina Zandonà, Luca Clivio, Luca Gabutti Published: 2025 ABSTRACT AbstractBackgroundHealthcare overuse is a widespread issue, accounting for approximately 10–30% of total healthcare costs. This study examines the impact of a computerized alert system designed to reduce inappropriate laboratory tests across five public hospitals in southern Switzerland.MethodsA 32-month observational study - between July 2021 and March 2024 - was conducted to assess the effectiveness of the alert system in discouraging unnecessary laboratory testing. Data were collected from five public hospitals where 12.4% of laboratory tests performed were identified as non-compliant with recommended minimal retesting intervals (MRI). Of these non-compliant tests, 11.4% were prevented by the alert system, indicating direct intervention efficacy.ResultsDistinct patterns in alert acceptance rates emerged over time. Hospital H4 showed a significant increase in alert acceptance, which may reflect contextual factors such as a work environment promoting open communication and kindness among healthcare professionals. Conversely, hospitals H1 and H5 experienced declining alert acceptance, likely due to habituation effects. A learning effect was suggested by a significant reduction in inappropriate tests for certain markers, such as C-Reactive Protein (CRP), over time. Following the alert’s implementation, the average rate of appropriate prescriptions across all hospitals was 87.5%. Specific tests, including PSA (19.4%), PTH (15.5%), ammonia (14.8%), and CRP (14.5%), had higher alert acceptance rates. The cost savings from avoided tests were estimated at EUR 1,642,417.ConclusionsThis study confirms that computerized alerts represent a gentle, effective and cost-efficient method to reducing overtreatment, increasing awareness among healthcare professionals, and supporting personalized medicine. The findings suggest: hypothetically - that organizational environments fostering human-centered care, effective communication and kindness may enhance the impact of alert systems, warranting further empirical investigation.What is known on this topic Overmedicalization accounts for 10–30% of healthcare costs, representing a significant concern in health resource management.The Minimum Repeat Interval (MRI) defines appropriate time at which a laboratory test can be repeated to minimize potentially unnecessary ones.Alerts in computerized medical records, indicating tests prescribed within the MRI, can effectively reduce unnecessary testing, especially when combined with other interventions. What this study adds This study demonstrates that laboratory alerts based on MRI effectively reduce overtreatment, enhancing personalized medicine, with unique data from 929,808 tests over 32 months.Alert compliance and appropriateness vary by test type, showing greater effectiveness for tests with higher risk of being inappropriately prescribed (e.g. CRP).Hypotheses are proposed regarding the potential influence of organizational culture - particularly environments promoting kindness - on the effectiveness of behavioral nudges. Keywords: Behavioral intervention, Clinical decision support system, Low-value care, Appropriateness, Organizational culture, Patient safety Introduction Healthcare overuse, also referred to as low-value care or overmedicalization, is a widespread phenomenon estimated at around 10–30% of all healthcare costs [1, 2]. It encompasses unnecessary medical practices, excessive diagnostic investigations, and inappropriate prescriptions of drugs or procedures [3–7]. In laboratory medicine, between 23% and 67% of tests are considered inappropriately prescribed [8], leading to patient discomfort, hospital-acquired anemia, prolonged hospital stay, and increased costs for the healthcare system [9–13]. Moreover, repeated blood sampling imposes an additional workload on nursing staff [14]. Reducing unnecessary testing therefore offers important benefits in terms of patient safety, staff workload, and cost containment [15–17]. Although the first Choosing Wisely recommendation to minimize unnecessary blood tests was published more than 10 years ago [18], the problem is not over. To guide retesting practices, the concept of a minimal retesting interval (MRI) has been introduced. MRIs define the recommended time frame during which repeating a laboratory test is unlikely to add new clinically relevant information and may instead increase the risk of false-positive results and overtreatment These intervals should be regarded as flexible guidelines rather than fixed thresholds, since specific patient circumstances may justify earlier retesting [19]. Over the past decade, several strategies have been implemented to reduce inappropriate laboratory testing, including educational interventions, audits, and computerized systems. A recent systematic review by Zare et al. [20] showed that clinical decision support systems (CDSS), particularly those using alerts or feedback algorithms, can significantly reduce inappropriate test ordering. However, these tools may lose effectiveness over time due to user habituation. From a behavioral perspective, computerized alerts can be conceptualized as nudges - gentle prompts that influence decision-making without restricting individual choice [21, 22]. Drawing on behavioral economics and cognitive psychology, nudges act by subtly modifying the choice architecture - for example, by providing timely reminders, framing default options, or highlighting key information - thereby facilitating evidence-based decisions [21, 23]. In healthcare, such interventions have been shown to reduce overuse and support adherence to guidelines while preserving clinician autonomy [24, 25]. In 2021, the Ente Ospedaliero Cantonale (EOC), a network of public hospitals in southern Switzerland, introduced a computerized alert system linked to MRI thresholds as part of its Smarter Medicine-Choosing Wisely campaign [26–28]. When a physician attempted to order a laboratory test earlier than the MRI, an alert appeared reminding them that a valid result was already available. Although all EOC hospitals share a centralized ICT infrastructure and similar technological standards, they differ in their management approaches and histories of quality improvement initiatives [29, 30]. The primary aim of this study was to assess whether the isolated introduction of a computerized alert system could significantly reduce potentially unnecessary laboratory tests and their associated costs. While the study was not initially designed to explore organizational determinants, previous research has shown that contextual features - such as workflow integration, leadership style, and organizational culture - may shape the effectiveness of behavioral interventions [24, 25, 31, 32]. In particular, hospitals that have long adopted Choosing Wisely-like approaches - emphasizing clinical appropriateness and the principle that “less is more” - may offer a more favorable environment for behavioral interventions aimed at reducing overuse [33–35]. Moreover, insights from the kindness-in-healthcare literature suggest that human-centred and compassionate organizational climates can enhance communication, engagement, and receptiveness to change, although their direct influence on behavioral interventions has not yet been empirically established [36]. Accordingly, the potential link between organizational culture - including Choosing Wisely orientation and kindness - and nudge effectiveness should be regarded as hypothesis-generating, providing a rationale for exploring contextual moderators in future research. We therefore aimed to answer the following research questions: Is the introduction of a computerized alert system effective, by itself, in reducing potentially inappropriate laboratory tests? Is the introduction of a computerized alert system effective, by itself, in reducing potentially inappropriate laboratory tests? Does the effectiveness of the alert remain stable over time, or does it diminish due to habituation? Does the effectiveness of the alert remain stable over time, or does it diminish due to habituation? Are there differences in alert effectiveness among hospitals within the same network? Are there differences in alert effectiveness among hospitals within the same network? Does the alert system reduce costs associated with laboratory testing? Does the alert system reduce costs associated with laboratory testing? Materials and methods This multicentric, nudging-based, single intervention observational study was conducted to monitor the percentage of appropriate and inappropriate prescriptions of 27 laboratory tests in five hospitals (H1-H5) of the Ente Ospedaliero Cantonale (EOC), Switzerland. The five hospitals share a centralized ICT service and innovation policybut differ in their organizational culture and management styles. In the EOC hospital network, laboratory tests are prescribed digitally via a software called PRESCO-LAB within the electronic patient record. Prescriptions can be carried out by doctors or other healthcare staff as an act of delegation by the physician. To reduce overprescribing, a computerized MRI-based alert system that generates pop-up alerts when predefined appropriateness criteria are violated, was introduced in the PRESCO-LAB software. The MRI for each laboratory test was defined through a structured peer discussion process among laboratory specialists and clinicians, following a review of the available literature [17, 37, 38]. In the absence of formal evidence-based guidelines, participants relied on published studies, clinical expertise, and patient safety considerations, generally favoring the shortest intervals judged to avoid risks to patients’ health. Agreement was reached through discussion until consensus was achieved, and the final decisions were documented in meeting minutes (Table 1). Table 1Minimal retesting interval (MRI): literature range and EOC choiceLaboratory testLiterature MRIEOC MRI1. Ammonia/14 days2. ANCA MPO qn6 months − 1 year [38]3 months3. ANCA PR3 qn6 months − 1 year [37, 38, 39]3 months4. anti-dsDNA qn3–6 months [38]3 months5. Prostate specific antigen [total PSA]6 weeks − 1 year [17, 37–41]6 months6. Cholesterol1–6 months [17, 37, 39, 42–51]1 month7. Complement C3c3 months [38]3 months8. Complement C43 months [38]3 months9. Protein electrophoresis7 days-3 months [17, 37, 50]3 months10. Glycated hemoglobin [HbA1c]40 days − 6 months[17, 37, 38, 42–44, 47, 48, 50–54]3 months11. Hepatitis B Ac anti-HBs1–3 months [38]1 months12. Ferritin7 days − 1 year[17, 37, 39, 43, 44, 47–49, 55]1 month13. Folic acid6 months − 1 year[17, 37, 49, 50, 56]3 months14. HDL- Cholesterol26 days-6 months[17, 37, 39, 42–45, 47, 50, 51]1 month15. IgG subclasses/1 year16. LDL-Cholesterol26 days − 6 months[17, 37, 42, 43, 45, 50, 51]1 month17. Intact parathyroid hormone [PTH]/3 months18. Natriuretic peptide [NT-proBNP]/14 days19. Procalcitonin [PCT]24 h[17, 37, 39, 43]7 days20. C-Reactive Protein [CRP]22 h − 7 days[17, 37, 39, 42, 43, 51, 55, 57–60]3 days21. Total proteins/3 months22. Reticulocytes/1 month23. Erythrocyte sedimentation/1 month24. Thyroid-stimulating hormone [TSH]4 weeks-3 years[17, 37, 39, 43, 44, 47, 48, 50, 55]2 months25. Urates/1 months26. Vitamin B 123 months − 1 year[17, 37, 39, 48–50, 56, 61]3 months27. Vitamin D [25-OH Vit D3]3 months − 1 year[37, 39, 43, 44, 48, 50, 56]6 months Minimal retesting interval (MRI): literature range and EOC choice 40 days − 6 months [17, 37, 38, 42–44, 47, 48, 50–54] 7 days − 1 year [17, 37, 39, 43, 44, 47–49, 55] 6 months − 1 year [17, 37, 49, 50, 56] 26 days-6 months [17, 37, 39, 42–45, 47, 50, 51] 26 days − 6 months [17, 37, 42, 43, 45, 50, 51] 24 h [17, 37, 39, 43] 22 h − 7 days [17, 37, 39, 42, 43, 51, 55, 57–60] 4 weeks-3 years [17, 37, 39, 43, 44, 47, 48, 50, 55] 3 months − 1 year [17, 37, 39, 48–50, 56, 61] 3 months − 1 year [37, 39, 43, 44, 48, 50, 56] To facilitate data analysis, some laboratory tests that are likely to be prescribed together were grouped (Table 2). Table 2The grouping of the laboratory tests under studyLaboratory testGrouping laboratory testsANCA (MPO e PR3)ImmunologyANCA MPO qnImmunologyANCA PR3 qnImmunologyanti-dsDNA qnImmunologyAmmoniaAmmoniaProstate specific antigenProstate specific antigenCholesterolCholesterolComplement C3cImmunologyComplement C4ImmunologyProtein electrophoresisElectrophoresisGlycated hemoglobin [HbA1c]Glycated hemoglobin [HbA1c]Hepatitis B Ac anti-HBsImmunologyFerritinSubstratesFolic acidSubstratesHDL- CholesterolCholesterolIgG subclassesElectrophoresisLDL- CholesterolCholesterolIntact parathyroid hormone [PTH]Intact parathyroid hormone [PTH]Natriuretic peptide [NT-proBNP]Natriuretic peptide [NT-proBNP]Procalcitonin [PCT]Procalcitonin [PCT]C-Reactive Protein [CRP]C-Reactive Protein [CRP]Total proteinElectrophoresisReticulocytesReticulocytesErythrocyte sedimentationErythrocyte sedimentationThyroid-stimulating hormone [TSH]Thyroid-stimulating hormone [TSH]UratesUratesVitamin B 12SubstratesVitamin D [25-OH Vit D3]Vitamin D The grouping of the laboratory tests under study The employee who prescribed the laboratory test was free to decide whether to accept or reject the alert; a gentle approach, which was preferred to a more coercive strategy. The effectiveness of the alert was monitored for 32 months (between July 2021 and March 2024). Over the entire period, the alert was the only strategy adopted to reduce the repetition of laboratory tests for both in- and out-patients. The extraction of the data from the local data warehouse was performed using user-defined and reproducible procedures, based on PowerBI in a fully anonymized mode. Data quality and coding accuracy were ensured through a reproducible workflow based on the hospital’s central data warehouse, developed using Microsoft Power BI. Data extraction was performed with a custom script that retrieved information from one of the developed ‘cubes’ (stored procedures) and generated a dataset ready for statistical analyses. The extracted dataset was manually reviewed to verify the absence of personal identifiers and to detect potential coding inconsistencies, thereby ensuring that the dataset provided to the statistician was fully anonymized and coded in line with the statistical protocol. The custom script was archived for future audits to document the extraction process and guarantee reproducibility. As the focus of the study was on the prescriptions themselves, no personal data of the patients have been collected. According to Swiss law, studies based solely on anonymous secondary data do not require ethics committee approval [62]. The extracted dataset consisted of 929,808 laboratory test prescriptions of five hospitals (H1, H2, H3, H4, H5). For each prescription, the dataset contained information on (a) the prescription date; (b) the hospital where the prescription was issued; (c) the laboratory test prescribed; (d) the triggering of the alert in case the prescription did not follow the recommended MRI; and (e) the outcome of the alert in case of triggering. The total monthly numbers of prescriptions, alerts ignored and accepted were then calculated overall, by location and by specific laboratory test. Descriptive statistics are expressed as relative frequencies and percentages for qualitative variables, while mean and standard deviation (SD) were used for quantitative variables after testing for normal distribution using the Kolmogorov-Smirnov test. Comparisons between hospitals regarding the total number of accepted alerts were performed using the chi-square test. A time series analysis was implemented to assess the trend in the percentage of adherence to the warning pop-up (i.e., accepted alerts) and the percentage of prescriptions complying with MRI recommendations (i.e., potentially appropriate prescriptions) during the observation period. The percentage of adherence to the pop-up warnings was calculated monthly as the ratio between the total number of accepted alerts and the total number of alerts that did not follow the recommended MRI. The percentage of potentially appropriate prescriptions was calculated monthly as the ratio between the total number of prescribed examinations that did not trigger a pop-up warning, and the total number of prescriptions registered during the considered period. After examining the data graphically to understand the basic characteristics of the time series data, to explain the trend over the period, the stationarity of the data was assessed with the Augmented Dickey-Fuller test and the most appropriate approach (i.e., linear vs. non-linear) was evaluated. The trend of the time series was then assessed using a linear regression analysis. This trend analysis made it possible to understand the long-term behavior of the data by indicating whether there was a statistically significant increase, decrease or stagnation over time. The trend analysis was performed overall, by hospital and by laboratory test. The assumptions of the linear regression models were checked by analyzing the residual plot and by the Durbin-Watson and Breusch-Pagan tests. The correct model specification was assessed using the regression specification error test. In addition, the stationarity of the result series and the residuals was tested using the extended Dickey-Fuller test (ADF). The normality of the residuals was checked using the Shapiro-Wilk test, and the presence of a serial correlation was further analyzed using the Breusch-Godfrey test and the autocorrelation function (ACF) of the residuals. The statistical significance was set at 0.05. All statistical analyses were performed with STATA18 (StataCorp, Collage Station, TX, USA). Results Overall, 929’808 laboratory tests were requested during the observation period; 115,248 out of 929,808 (12.4%) laboratory test requests violated the appropriateness criteria defined by the MRI and triggered the appearance of the electronic alert (warning pop-up). After the alert appeared, 13,698 requests were canceled (1.5% of the total requested laboratory tests and 11.9% of the laboratory tests with alert). Percentages of accepted alerts above 10% were observed for PSA (19.4%), PTH (15.5%), Ammonia (14.8%), CRP (14.5%), Vitamin D (12.7%), Urates (12.5%), Reticulocytes (12.2%), and Procalcitonin (11.2%) (Fig. 1). Percentage of laboratory test with no alert, ignored alert or accepted alerts. Percentage of laboratory test with no alert (blue), ignored alert (orange) or accepted alerts (grey) after implementation of the computerized alert system Higher percentages of accepted alerts were observed in H4 (15.0%) and H3 (14.3%) compared to H2 (11.8%), H1 (10.9%), and H5 (10.3%) (p < 0.001). Overall, the percentage of accepted alerts significantly decreased over the observed period (β = -0.08, SE: 0.02, p = 0.003, Fig. 2a), while no statistically significant change emerged about the potentially appropriate prescriptions (Fig. 2b). For percentage of accepted alerts, the series was stationary (ADF: Z = − 3.419, p = 0.010). Model residuals were approximately normal (Shapiro-Wilk W = 0.986, p = 0.937), homoscedastic (Breusch-Pagan χ² [1] = 0.77, p = 0.380), and stationary (ADF: Z = − 4.693, p = 0.0001). No residual autocorrelation was detected (Breusch-Godfrey χ² [1] = 0.112, p = 0.738). For potentially appropriate prescriptions, the series was stationary (ADF: Z = − 4.044, p = 0.0012). Model residuals were approximately normal (Shapiro-Wilk W = 0.972, p = 0.526), homoscedastic (Breusch-Pagan χ² [1] = 0.36, p = 0.547), and stationary (ADF: Z = − 4.149, p = 0.0008). No residual autocorrelation was detected (Breusch-Godfrey χ² [1] = 1.495, p = 0.221). Percentage of accepted alerts and Percentage of potentially appropriate prescriptions. (a) Percentage of accepted alerts. (b) Percentage of potentially appropriate prescriptions. The green dashed line represents the actual monthly data. The orange line represents the linear trend as calculated after regression analysis. The shadow area represents the 95% confidence interval The monthly percentage of accepted alerts decreased significantly in H1 (β = -0.22, SE: 0.03, p < 0.001, Fig. 3a) and H5 (β = − 0.13, SE: 0.04 p = 0.002, Fig. 3e), while bucking the overall trend, the monthly percentage of accepted alerts in H4 increased significantly (β = 0.16, SE: 0.07, p = 0.020, Fig. 3d). In H2 and H3 no statistically significant trends emerged (Fig. 3b and c). Percentages of accepted alerts by hospital: (a) H1, (b) H2, (c) H3, (d) H4, (e) H5. The green dashed line represents the actual monthly data. The orange line represents the linear trend as calculated after regression analysis. The shadow area represents the 95% confidence interval The percentage of potentially appropriate prescriptions significantly increased in H4 (β = 0.17, SE: 0.03, p < 0.001, Fig. 4d) and significantly decreased in H1 (β = -0.11, SE: 0.03, p < 0.001, Fig. 4a), while no statistically significant changes emerged in the other hospitals (Fig. 4b, c and e). Percentages of potentially appropriate prescriptions by hospital: a) H1, b) H2, c) H3, d) H4, e) H5. The green dashed line represents the actual monthly data. The orange line represents the linear trend as calculated after regression analysis. The shadow area represents the 95% confidence interval The monthly percentage of accepted alerts showed significant decreasing trends for several tests, including inflammatory markers (e.g. C-reactive protein, erythrocyte sedimentation rate, procalcitonin) and selected biochemical measures (e.g. ammonia, vitamin D, urates). In contrast, the percentage of potentially appropriate prescriptions increased over time for some tests (notably PSA, cholesterol, CRP and ESR), while decreasing for others (e.g. electrophoresis, HbA1c, natriuretic peptide and reticulocytes). The complete results with coefficients and standard errors are shown in Table 3. Diagnostic checks were systematically performed in all subgroup analyses. In most models, the assumptions were met without relevant concerns. Seven subgroups showed violations, mainly residual non-normality or autocorrelation. In these cases, inference was repeated using heteroscedasticity- and autocorrelation-consistent (HAC/Newey-West) standard errors, together with sensitivity analyses based on Prais-Winsten AR [1] models and residual bootstrap. The results remained consistent even with robustness checks, which underpins the reliability of the reported results. Table 3Trends in percentage of accepted alerts and potentially appropriate prescriptions by exam% of potentially appropriate prescriptions: Coeff. (Standard Error)% of accepted alerts: Coeff. (Standard Error)Ammonia0.08 (0.10)-0.44** (0.20)C-Reactive Protein0.08*** (0.01)-0.06*** (0.01)Electrophoresis*-0.10** (0.05)-0.33*** (0.11)Erythrocyte sedimentation0.04*** (0.01)-0.30** (0.13)Glycated haemoglobin-0.13*** (0.04)0.15 (0.15)Immunology**-0.06*** (0.02)-0.19 (0.23)Natriuretic peptide-0.06** (0.03)-0.12 (0.09)Procalcitonin0.10 (0.06)-0.38** (0.15)Prostate specific antigen0.35** (0.06)-0.05 (0.05)PTH-0.27 (0.13)-0.29** (0.14)Reticulocytes-0.51*** (0.05)0.08 (0.09)Substrates***-0.08 (0.05)-0.10 (0.08)Thyroid-stimulating hormone0.02 (0.03)0.10 (0.08)Cholesterol0.07** (0.03)-0.21 (0.15)Urates0.07 (0.04)-0.20** (0.08)Vitamin D-0.06 (0.05)-0.24** (0.09)Legend: *Electrophoresis includes Protein electrophoresis, IgG subclasses and Total proteins. **Immunology includes ANCA MPO qn, ANCA PR3 qn, Complement C3c, Complement C4, Hepatitis B Ac anti-HBs, anti-dsDNA qn. ***Substrates include Ferritin, Folic Acid, Vitamin B 12 Trends in percentage of accepted alerts and potentially appropriate prescriptions by exam Legend: *Electrophoresis includes Protein electrophoresis, IgG subclasses and Total proteins. **Immunology includes ANCA MPO qn, ANCA PR3 qn, Complement C3c, Complement C4, Hepatitis B Ac anti-HBs, anti-dsDNA qn. ***Substrates include Ferritin, Folic Acid, Vitamin B 12 The total potential financial savings of the withdrawn tests amounted to EUR 1,642,417 over the entire study period (Table 4). Table 4Economic impact of to the implementation of the computerized alert systemLaboratory testEOC MRIN. tests*N. accepted alertTest cost (EUR)Total tests cost (EUR)Cost for n° accepted allert (EUR)%Ammonia14 days1’46130139.958’29412’01021%ANCA MPO qn3 months8014326.621’3071’1445%ANCA PR3 qn3 months6603426.617’5569045%anti-dsDNA qn3 months1’4746649.572’9043’2644%Prostate specific antigen [total PSA]6 months17’7894’78011.2199’41553’58427%Cholesterol1 month32’3811’3152.478’3623’1824%Complement C3c3 months3’08620021.967’5834’3806%Complement C43 months3’23016921.970’7373’7015%Protein electrophoresis3 months13’95387829.5411’61425’9016%Glycatws hemoglobin [HbA1c]3 months21’3121’19416.9360’38620’1916%Hepatitis B Ac anti-HBs1 month10’32430219.0196’3625’7443%Ferritin1 month48’8673’8147.5365’52528’5298%Folic acid3 months33’0703’78812.5412’38347’23611%HDL- Cholesterol1 month30’3461’2068.4253’38910’0704%IgG subclasses1 year1’92239991.3175’51736’43721%LDL-Cholesterol1 month31’2791’2763.8118’2354’8234%Intact parathyroid hormone [PTH]3 months14’1343’00635.2497’517105’81121%Natriuretic peptide [NT-proBNP]14 days17’4321’78366.61’160’797118’73010%Procalcitonin [PCT]7 days6’9391’00879.9554’42680’53915%C-Reactive Protein [CRP]3 days348’04964’6309.53’309’946614’63119%Total proteins3 months42’8014’8192.4104’00611’71011%Reticulocytes1 month24’7114’07612.6311’11151’31716%Erythrocyte sedimentation1 month37’3122’6241.035’4462’4937%Thyroid-stimulating hormone [TSH]2 months19’6212’0998.5167’17117’88311%Urates1 month48’8299’0922.6128’90924’00319%Vitamin B 123 months33’9334’13223.7802’85597’76312%Vitamin D [25-OH Vit D3]6 months28’8685’08750.41’455’236256’43618% TOTAL 874’584 122’121 681.2 11’406’989 1’642’417 14% Economic impact of to the implementation of the computerized alert system Alert was implemented by the EOC’s IT department and the internal IT development costs were about EUR 55,000. To these must be added the project costs, which are however part of the ongoing quality management program. Discussion The inappropriate prescription of laboratory tests is a phenomenon that not only leads to patient discomfort, risk of hospital-acquired anemia and prolonged hospital stay, but also to a waste of resources by the healthcare system. Several studies investigated this phenomenon and proposed methods that can effectively reduce unnecessary repetition of laboratory tests. These include the use of computer alerts, the avoidance of clinical circumstance-oriented panels of tests, changes in hospital policy (e.g. workflow oriented) and the introduction of audits and educational interventions. The appropriateness of laboratory tests is usually defined by the MRI concept, i.e. the minimum time that should elapse before a test is repeated for a given pathological condition. However, different MRIs have been proposed for the same laboratory test and no agreement has yet been reached. This study aimed to analyze the effectiveness of a single intervention, namely the introduction of computer alerts for nudging, to curb the phenomenon of inappropriate prescriptions of laboratory tests in a network of five hospitals in southern Switzerland. Specifically, we wanted to examine the effectiveness of the alerts over time, in terms of an increase/decrease in appropriateness (i.e., compliance with the MRI), and the possible habituation effect (i.e., significant decrease in compliance due to the habituation to alerts potentially perceived as disturbing). We also wanted to evaluate possible differences among the five hospitals to assess consistency with the literature on nudging strategies and the potential impact of a kindness oriented organizational culture. The study was conducted by monitoring the number of accepted/rejected alerts for 27 laboratory tests over a period of 32 months (July 2021-March 2024). During this period, 12.4% of the tests were judged as potentially inappropriate based on MRI. The average rate of laboratory tests performed in accordance with the MRI was therefore 87.6%, a value in line with the literature (Bates et al..: 96.38%; Bridges at al.: 97.4%; Lippi et al.: 78%; Niès et al..: 84.5%; Rudolf et al..: 48.5%). The great heterogeneity among the studies (i.e., different MRIs, sample sizes, study designs and time intervals) must however be considered in the comparison. In some laboratory tests (i.e., CRP, Erythrocyte sedimentation rate, PSA, Total Cholesterol), a significant decrease in the percentage of inappropriateness was observed over time, a phenomenon that suggests a change in prescribing strategy, probably due to an educational effect of the alert; while the appropriateness of other laboratory tests (Procalcitonin, PTH, Thyroid-stimulating hormone) was not affected by the alert introduction. The improvement in CRP appropriateness may reflect previous awareness campaigns in our hospital network [13]; reactivated by the alerts. This highlights how contextual factors beyond the intervention itself may influence effectiveness. A favorable trend was also observed for Erythrocyte sedimentation rate, PSA and Total Cholesterol. These laboratory tests do not have major quantitative impact at the network level, but it is easy for the prescriber to guess the reasons for the MRI, given the low tendency for the value to change in a short time lapse. For other laboratory tests, whose clinical significance requires a more complex interpretation, such as Immunology, Protein electrophoresis, BNP, Glycated hemoglobin, and Reticulocytes, prescriptive appropriateness decreased significantly over the observation period, but without habituation to the alert except for Protein electrophoresis. Thus, it can be said that the alert had an overall positive impact on prescriptive behaviors, especially for laboratory tests where the futility of short-term repetition is easily recognizable. Beyond statistical significance, the clinical relevance of reducing unnecessary tests deserves emphasis. Unwarranted laboratory testing not only generates waste but can trigger diagnostic and therapeutic cascades following meaningless fluctuations in results. Such variations - though clinically irrelevant - may be interpreted as signs of disease progression and lead to unnecessary treatment adjustments or further investigations. For instance, repeating C-reactive protein (CRP) measurements within a short interval can prolong intravenous antibiotic therapy without improving patient outcomes [13], while inappropriate repetition of ammonia testing may reinforce the misconception that biochemical values, rather than clinical status, define treatment response in hepatic encephalopathy. As highlighted by the Choosing Wisely Italia recommendations, elevated ammonia levels alone do not aid diagnosis or management of hepatic encephalopathy in patients with chronic liver disease and rarely correlate with severity or outcome [63–66]. Similarly, spurious fluctuations in PSA or PTH values may lead clinicians to misinterpret benign variability as a trend requiring intervention, prompting unnecessary testing or overtreatment. Reducing inappropriate laboratory requests therefore contributes not only to economic efficiency but also to clinical safety, by minimizing avoidable procedures, patient anxiety, and exposure to inappropriate therapies. These effects, though not captured by statistical measures alone, represent meaningful improvements in the quality and appropriateness of care. In 11.9% of cases, the alert was accepted and the prescription cancelled, which corresponds to laboratory tests that would probably have been unnecessary. Overall, a significant decrease in accepted alerts was observed during the study period. This suggests a component of habituation and possibly alert fatigue, a well-documented phenomenon in the literature on clinical decision support systems [67]. The only exception to this trend concerns H4, the hospital where the alerts appeared to be most effective. In fact, it was the hospital with the highest percentage of alert acceptance (15.0%) and the one in which the percentage of appropriate prescriptions increased significantly over time, showing a learning effect in relation to MRI. One possible contextual explanation relates to the organizational culture of Hospital H4. H4 has long promoted a people-centered culture [29, 68] and, since 2021, top management has deliberately fostered an environment of kindness among staff, patients, caregivers, and the wider care community. These initiatives led to the hospital being officially recognized as the “Ambassador of Kindness” for Switzerland by the International Kindness Movement. Activities have included training in complaint management, creation of educational and promotional materials (videos, posters, a “kindness advent calendar”), staff well-being initiatives, public events on World Kindness Day, and participatory meetings with patients and families. Other hospitals within the EOC network have progressively adopted similar projects since 2024. H4 was also the first hospital in Southern Switzerland to establish a dedicated quality and patient safety service in 1995 [30] and has long been considered by the EOC a pilot site for testing innovative management tools and approaches [69–71]. These characteristics may have contributed to an organizational climate potentially more open to behavioral change, which could help explain the higher responsiveness to the alert intervention. While our study was not designed to test causal mechanisms, it is plausible - though unproven - that hospitals with leadership styles and organizational cultures emphasizing human-centeredness and kindness may be more receptive to behavioral nudges. This interpretation aligns with prior work suggesting that contextual and cultural factors can influence the adoption and effectiveness of behavioral interventions in healthcare [27, 30, 31] and with recent evidence highlighting how compassionate and people-centered organizational climates may facilitate positive behavioral change [36]. Nevertheless, the potential link between kindness, organizational culture, and the effectiveness of nudging interventions should be regarded as hypothesis-generating and warrants further empirical investigation. In terms of the impact of the alerts on efficiency we are aware that in a hospital system with integrated laboratory medicine fewer laboratory tests do not automatically mean lower costs, but we believe that some costs, particularly for reagents, have indeed been saved. Although overall compliance with the nudge strategy described in this work was not impressive, the cumulative financial impact calculated using the costs of the test charged to the services was substantial (more than EUR 1.6 million during the study period, representing 14% of the total cost of the laboratory tests considered in the study). The estimated savings of EUR 1.6 million likely represent a conservative figure, as additional downstream costs (e.g., follow-up investigations triggered by false-positive results) were not included. The internal cost of implementing the alert (approximately EUR 60,000 in total) being reasonable. To achieve even greater cost savings and educate prescribers to use laboratory resources more appropriately and efficiently it would be useful to combine the computerized alert system with other strategies. These include training on Choosing Wisely recommendations and planetary health requirements, as well as detailing the reason for introducing the alert system, scheduling regular discussions on compliance data and adapting hospital policies on the subject. The findings of this study suggest that a computerized alerting system is a cost-effective and undemanding solution for hospital management, with implications for policy. Such an intervention, however, is not very effective as the sole strategy to limit inappropriate laboratory tests and is subject to habituation. Finally, such a nudging appears to be most effective in hospitals with an organizational culture that puts people at the center by promoting kindness [25, 72, 73]. The study presents some limitations. First, the choice of MRIs was not based on established evidence-based guidelines, as there is no consensus in the literature. Our hospital network therefore developed an internal consensus through a structured peer discussion, taking into account available evidence, clinical judgment, and patient safety considerations. Although the process was documented, the selected time intervals may not be appropriate for every pathological condition. Second, the potential positive effects on staff in terms of awareness, knowledge, and behavioral change were not quantified. Third, possible confounding factors, such as other ongoing quality improvement initiatives, may have influenced the observed outcomes. Fourth, we did not assess clinical outcomes such as patient morbidity or mortality, which limits the ability to determine the safety implications of reduced testing. Finally, the generalisability of our findings beyond the Swiss hospital context remains uncertain, given differences in healthcare systems and organizational cultures. In terms of strengths, the large number of observations (929,808 laboratory tests), the monitoring duration (32 months), the number of laboratory tests (n = 27), and the number of alerts (115,248) are unique in the literature. The design focused on a single intervention (introduction of an alert) compared to the implementation of a mix of strategies (e.g., staff training, change in hospital policy, etc.), also increases the reliability of the results. Conclusions This 32-month multicentric Swiss study provides a detailed analysis showing that the introduction of an electronic alert system to reduce overtreatment may operate differently depending on the context in which it is applied. Overall, a habituation effect was observed; however, the effectiveness of such interventions appears to be influenced by the organizational culture. The findings suggest that alert systems may be more effective in hospitals fostering a culture of open communication, kindness and human-centred care. The use of an alert system has different effects on the prescription according to the laboratory test: the habituation effect occurs for many, but not for all (e.g. CRP), and the same is true for the presence of a learning effect. This could be because the alert system functions as an advice, prompting professionals to re-assess the actual necessity to repeat the laboratory test according to the patient’s medical needs. This type of decision-making tool can reduce overtreatment and support the healthcare system in developing personalized medicine. To achieve more effective results, healthcare managers should consider implementing this kind of intervention alongside others, particularly educational interventions and monitoring systems. Finally, whether combining the Choosing Wisely philosophy with cultural changes that create a more favorable environment for adopting recommendations (such as leadership-driven promotion of open communication and kindness) leads to greater compliance with appropriateness of care remains to be confirmed in future studies, and should be regarded as a hypothesis-generating observation rather than a demonstrated relationship. Acknowledgements The authors thank the following colleagues (in alphabetical order): Prof. Marco Bertoni and Prof. Luca Corazzini, with whom we initiated preliminary research on the appropriateness of laboratory tests, for engaging discussions; Dr. med Nicolò Saverio Centemero and Brigitte Waldispühl Suter, Computerised Clinical Processes EOC Service, for coordinating the Alert Introduction; Dr. Roberto della Bruna, Head of the Clinical Chemistry Service, Deputy Director and Head of the Quality Sector at the Institute of Laboratory Medicine, EOC, and Alessia Ricciardi, Junior Controller at Locarno Hospital (EOC), for providing information on the costs of the selected laboratory tests; Prof. Dr. med Paolo Ferrari, Head of Medical Area DG EOC, for participating in the design and approving the Alert Introduction Project; Prof. Giuseppe Lippi, Full Professor of Clinical Biochemistry, Dean of Medical Faculty and Director of Laboratory Medicine Service at University Hospital of Verona, for the exchange of information in relation to the Research Project on the topic, he published in 2015; Alessandro Merler, Head of Business Intelligence EOC, and Federica Fiorentini, for providing data on laboratory tests and the alert, and Lorenzo Ruinelli, from the EOC ICT service, for supporting the project; and all the clinicians of the EOC Hospitals for their daily commitment to providing quality care. Without the latter, this project would not have existed. Abbreviations Clinical decision support systems C-reactive protein Ente Ospedaliero Cantonale Glycated hemoglobin Hospitals of the EOC network Information and communication technology Minimal retesting interval Natriuretic peptide Procalcitonin Prostate specific antigen Intact parathyroid hormone Thyroid-stimulating hormone Author contributions AG e LG conceptualized the study aims and design. AG drafted the original manuscript and wrote the main manuscript text. AG, MZ, MLG and LG contributed to the revision of the original draft. LC extracted the study data; MLG performed the statistical analysis and AG the calculations for the financial part. AG and LG ensured the general supervision. All authors read and approved the final manuscript. Funding This research received no specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Data availability The data and materials that support the findings of this study are available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The data and materials that support the findings of this study are available from the corresponding author upon reasonable request.