Michael M. Havranek

Hospital readmissions prolong patient suffering and increase healthcare expenditures. Unplanned readmission rates, such as those reported by the Center for Medicare & Medicaid Services (CMS), distinguish between planned and unplanned readmissions. However, within unplanned readmissions, there is no distinction between those that are preventable versus unpreventable by the hospitals. Alternative approaches attempting to identify potentially preventable readmissions directly from coded medical data have been explored but have shown low sensitivity. Consequently, identifying preventable readmissions remains a time-consuming task for hospital quality managers seeking to allocate improvement resources effectively. To address this challenge, we aimed to develop and evaluate a probabilistic approach to improve the identification of preventable readmissions among unplanned readmissions. Using a retrospective record review of 600 single inpatient stays from a tertiary referral hospital group in Switzerland, we investigated the hypothesis that readmitted patients with a low expected probability for readmission (based on a logistic regression model using patient characteristics) would retrospectively show higher odds of having experienced a potentially or most likely preventable readmission (as assessed by the reviewers). The results confirmed our hypothesis: patients in the third with the lowest expected probability of readmission (compared with those in the highest third) had 6.6 and 8.7 times higher odds, respectively, of having experienced a potentially or most likely preventable readmission. Among preventable readmissions, the leading causes of readmission were surgical complications, medication-related reasons, nonsurgical complications, and premature discharge. Our proposed probabilistic approach can be used by hospital quality managers to focus case finding efforts on unexpectedly readmitted patients and aid effective resource allocation for improvement initiatives.

Hospital readmissions prolong patient suffering and increase healthcare expenditures. While several studies have attempted to develop prediction models to reduce readmissions, most have demonstrated modest predictive accuracy. To improve upon prior approaches, we conducted an overview of systematic reviews to identify the most relevant predictor variables, then subsequently developed machine learning models in a retrospective, multisite study across eight hospitals. The patient sample comprised 200,799 inpatient stays from eligible hospitalizations, based on the Centers for Medicare and Medicaid Services (CMS) definition of unplanned readmissions within 30 days of discharge. We constructed random forest models and evaluated out-of-sample performance using the area under the receiver operating characteristic curve (AUC) across different train–test splits. The hospital-wide sample was divided into medical and surgical cohorts to investigate predictor importance across different patient populations. The average AUC score was 0.78 ± 0.01 (mean ± standard deviation [SD]). Patients’ diagnoses were the most important predictor variables (contributing 18.4% ± 0.15 to the model’s decision, mean ± standard error [SE]), followed by nursing assessments (11.2% ± 0.04, mean ± SE) and procedural information (10.8% ± 0.09, mean ± SE). Comparing medical and surgical patients, we found that medications and prior healthcare use (e.g., prior emergency encounters) were more important in the medical compared with the surgical cohort, whereas procedural information and healthcare provider information (e.g., physician caseload) were more relevant in the surgical relative to the medical cohort. In conclusion, we have established the feasibility of using Swiss electronic medical record (EMR) data to accurately predict unplanned readmissions. The reported variable importances may guide future research and inform development of clinical decision support systems aimed at reducing readmissions.

Hypoglycemia is a major challenge for people with diabetes. Therefore, glycemic monitoring is an important aspect of diabetes management. However, current methods such as finger pricking and continuous glucose monitoring systems (CGMS) are invasive, and hypoglycemia has still been shown to occur despite advancements in CGMS. Consequently, a growing body of research has been directed toward noninvasive hypoglycemia detection, relying on data from medical devices and wearables that can record physiological changes elicited by hypoglycemia. Consumer-grade wearables such as smartwatches remain an attractive yet underexplored candidate for such applications. Therefore, we explored the potential of a consumer-grade wearable for hypoglycemia prediction and investigated differing feature importance during waking and sleeping times. Smartwatch data from 18 adults with type 1 diabetes was collected, preprocessed, and imputed. Machine learning (ML) models were built using a tree-based ensemble algorithm to detect hypoglycemic events registered by CGMS. Models were built in a personalized manner using the same participant’s data for training and testing, with separate modeling for daytime and nighttime. The relative importance of input features on model decisions was analyzed using SHAP (SHapley Additive exPlanations). Seventeen personalized models were built with an average area under the receiver operating characteristic curve (AUROC) score of 0.74 ± 0.08. Average specificity and sensitivity were 0.76 ± 0.18 and 0.71 ± 0.15, respectively. Time-of-day, activity, and cardiac features showed comparable importance in daytime models (29.9%, 28.5%, and 24%, respectively), while in nighttime models, cardiac features demonstrated the highest importance (42.2%) followed by time-of-day features (37.5%) and respiratory features (15.2%). In summary, we demonstrate the potential of consumer-grade wearables in noninvasive hypoglycemia detection. By additionally considering different physiological states (waking and sleeping) during modeling, our results offer further insights into differences in relative feature importance influencing the model’s decision, guiding future research in this area.

Background: Hospital readmission rates are used for quality and pay‐for‐performance initiatives. To identify readmissions from administrative data, two commonly employed methods are focusing either on unplanned readmissions (used by the Centers for Medicare & Medicaid Services, CMS) or potentially avoidable readmissions (used by commercial vendors such as SQLape or 3 M). However, it is not known which of these methods has higher criterion validity and can more accurately identify actually avoidable readmissions.Objectives: A manual record review based on data from seven hospitals was used to compare the validity of the methods by CMS and SQLape.Methods: Seven independent reviewers reviewed 738 single inpatient stays. The sensitivity, specificity, positive predictive value (PPV), and F1 score were examined to characterize the ability of an original CMS method, an adapted version of the CMS method, and the SQLape method to identify unplanned, potentially avoidable, and actually avoidable readmissions.Results: Both versions of the CMS method had greater sensitivity (92/86% vs. 62%) and a higher PPV (84/91% vs. 71%) than the SQLape method, in terms of identifying their outcomes of interest (unplanned vs. potentially avoidable readmissions, respectively). To distinguish actually avoidable readmissions, the two versions of the CMS method again displayed higher sensitivity (90/85% vs. 66%), although the PPV did not differ significantly between the different methods.Conclusions: Thus, the CMS method has both higher criterion validity and greater sensitivity for identifying actually avoidable readmissions, compared with the SQLape method. Consequently, the CMS method should primarily be used for quality initiatives.

There currently exists no comprehensive and up-to date overview on the financial impact of the different adverse events covered by the Patient Safety Indicators (PSIs) from the Agency for Healthcare Research and Quality. We conducted a retrospective case-control study using propensity score matching on a national administrative data set of 1 million inpatients in Switzerland to compare excess costs associated with 16 different adverse events both individually and on a nationally aggregated level. After matching 8,986 cases with adverse events across the investigated PSIs to 26,931 controls, we used regression analyses to determine the excess costs associated with the adverse events and to control for other cost-related influences. The average excess costs associated with the PSI-related adverse events ranged from CHF 1,211 (PSI 18, obstetric trauma with instrument) to CHF 137,967 (PSI 10, postoperative acute kidney injuries) with an average of CHF 27,409 across all PSIs. In addition, adverse events were associated with 7.8-day longer stays, 2.5 times more early readmissions (within 18 days), and 4.1 times higher mortality rates on average. At a national level, the PSIs were associated with CHF 347 million higher inpatient costs in 2019, which corresponds to about 2.2% of the annual inpatient costs in Switzerland. By comparing the excess costs of different PSIs on a nationally aggregated level, we offer a financial perspective on the implications of in-hospital adverse events and provide recommendations for policymakers regarding specific investments in patient safety to reduce costs and suffering.

Background: Despite the established positive association between patient experience and patient volume, the relationship between patient experience and the financial performance of hospitals has not been studied thoroughly. Methods: To investigate this relationship, we used longitudinal data from 132 Swiss acute-care hospitals from 2016 to 2019 to examine the associations between patient experience and the proportion of elective patients, revenue, costs, and profits of hospitals. To account for a potential time lag effect, we utilized annual patient experience data and employed multilevel mixed-effects regression modeling to investigate its association with the aforementioned financial performance indicators for the following year. Results: Data for private and public hospitals were analyzed both separately and in combination, to account for the different proportions of elective patients in these types of hospitals. The resulting mixed models, revealed that for each year studied, the previous year’s patient experience was positively associated with the current year’s proportion of elective patients (β = 0.09, p = 0.004, all hospitals) and revenue (β = 1789.83, p = 0.037, private hospitals only), and negatively associated with costs (β = − 1191.13, p = 0.017, all hospitals); but not significantly associated with future profits (β = 629.12, p = 0.240, all hospitals). Conclusions: This analysis showed that better patient experience is associated with a higher proportion of elective patients, greater revenue, and lower costs. Our findings may assist hospital managers and regulators in identifying strategies to increase revenue and reduce costs.

The validity of the Agency for Healthcare Research and Quality’s Patient Safety Indicators (PSIs) has been established in the USA and Canada. However, these indicators are also used for hospital benchmarking and cross-country comparisons in other nations with different health-care settings and coding systems as well as missing present on admission (POA) flags in the administrative data. This study sought to comprehensively assess and compare the validity of 16 PSIs in Switzerland, where they have not been previously applied. We performed a medical record review using administrative and electronic medical record data from nine Swiss hospitals. Seven independent reviewers evaluated 1245 cases at various hospitals using retrospective data from the years 2014–18. True positives, false positives, positive predictive values (PPVs), and reasons for misclassification were compared across all investigated PSIs, and the documentation quality of the PSIs was examined. PSIs 6 (iatrogenic pneumothorax), 10 (postoperative acute kidney injury), 11 (postoperative respiratory failure), 13 (postoperative sepsis), 14 (wound dehiscence), 17 (birth trauma), and 18 and 19 (obstetric trauma with or without instrument) showed high PPVs (range: 90–99%) and were not strongly influenced by missing POA information. In contrast, PSIs 3 (pressure ulcer), 5 (retained surgical item), 7 (central venous catheter-related bloodstream infection), 8 (fall with hip fracture), and 15 (accidental puncture/laceration) showed low PPVs (range: 18–49%). In the case of PSIs 3, 8, and 12 (perioperative embolism/thrombosis), the low PPVs were largely due to the lack of POA information. Additionally, it was found that the documentation of PSI 3 in discharge letters could be improved. We found large differences in validity across the 16 PSIs in Switzerland. These results can guide policymakers in Switzerland and comparable health-care systems in selecting and prioritizing suitable PSIs for quality initiatives. Furthermore, the national introduction of a POA flag would allow for the inclusion of additional PSIs in quality monitoring.

Structural factors can influence hospital costs beyond case-mix differences. However, accepted measures on how to distinguish hospitals with regard to cost-related organizational and regional differences are lacking in Switzerland. Therefore, the objective of this study was to identify and assess a comprehensive set of hospital attributes in relation to average case-mix adjusted costs of hospitals. Using detailed hospital and patient-level data enriched with regional information, we derived a list of 23 cost predictors, examined how they are associated with costs, each other, and with different hospital types, and identified principal components within them. Our results showed that attributes describing size, complexity, and teaching-intensity of hospitals (number of beds, discharges, departments, and rate of residents) were positively related to costs and showed the largest values in university (i.e., academic teaching) and central general hospitals. Attributes related to rarity and financial risk of patient mix (ratio of rare DRGs, ratio of children, and expected loss potential based on DRG mix) were positively associated with costs and showed the largest values in children’s and university hospitals. Attributes characterizing the provision of essential healthcare functions in the service area (ratio of emergency/ ambulance admissions, admissions during weekends/ nights, and admissions from nursing homes) were positively related to costs and showed the largest values in central and regional general hospitals. Regional attributes describing the location of hospitals in large agglomerations (in contrast to smaller agglomerations and rural areas) were positively associated with costs and showed the largest values in university hospitals. Furthermore, the four principal components identified within the hospital attributes fully explained the observed cost variations across different hospital types. These uncovered relationships may serve as a foundation for objectifying discussions about cost-related heterogeneity in Swiss hospitals and support policymakers to include structural characteristics into cost benchmarking and hospital reimbursement.

Social anxiety disorder (SAD) is the most common anxiety disorder and has considerable negative impact on social functioning, quality of life, and career progression of those affected. Gelotophobia (the fear of being laughed at) shares many similarities and has therefore been proposed as a subtype of SAD. This hypothesis has, however, never been tested in a clinical sample. Thus, the relationship between gelotophobia, SAD and avoidant personality disorder (APD) was investigated by examining a sample of 133 participants (64 psychiatric patients and 69 healthy controls matched for age and sex) using the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders (4th edition) and an established rating instrument for gelotophobia (GELOPH). As expected, gelotophobia scores and the number of gelotophobic individuals were significantly higher among patients with SAD (n = 22) and APD (n = 12) compared to healthy controls and other psychiatric patients. Furthermore, gelotophobia scores were highest in patients suffering from both SAD and APD. In fact, all patients suffering from both disorders were also suffering from gelotophobia. As explained in the discussion, the observed data did not suggest that gelotophobia is a subtype of SAD. The findings rather imply that the fear of being laughed at is a symptom characteristic for both SAD and APD. Based on that, gelotophobia may prove to be a valuable additional diagnostic criterion for SAD and APD and the present results also contribute to the ongoing debate on the relationship between SAD and APD.