Targeted learning in observational studies with multi-valued treatments: An evaluation of antipsychotic drug treatment safety.

We investigate estimation of causal effects of multiple competing (multi-valued) treatments in the absence of randomization. Our work is motivated by an intention-to-treat study of the relative cardiometabolic risk of assignment to one of six commonly prescribed antipsychotic drugs in a cohort of nearly 39 000 adults with serious mental illnesses. Doubly-robust estimators, such as targeted minimum loss-based estimation (TMLE), require correct specification of either the treatment model or outcome model to ensure consistent estimation; however, common TMLE implementations estimate treatment probabilities using multiple binomial regressions rather than multinomial regression. We implement a TMLE estimator that uses multinomial treatment assignment and ensemble machine learning to estimate average treatment effects. Our multinomial implementation improves coverage, but does not necessarily reduce bias, relative to the binomial implementation in simulation experiments with varying treatment propensity overlap and event rates. Evaluating the causal effects of the antipsychotics on 3-year diabetes risk or death, we find a safety benefit of moving from a second-generation drug considered among the safest of the second-generation drugs to an infrequently prescribed first-generation drug known for having low cardiometabolic risk.

Healthcare Access for a Diverse Population with Schizophrenia Following the Onset of the COVID-19 Pandemic.

COVID-19 has had a disproportionate impact on the most disadvantaged members of society, including minorities and those with disabling chronic illnesses such as schizophrenia. We examined the pandemic's impacts among New York State's Medicaid beneficiaries with schizophrenia in the immediate post-pandemic surge period, with a focus on equity of access to critical healthcare. We compared changes in utilization of key behavioral health outpatient services and inpatient services for life-threatening conditions between the pre-pandemic and surge periods for White and non-White beneficiaries. We found racial and ethnic differences across all outcomes, with most differences stable over time. The exception was pneumonia admissions-while no differences existed in the pre-pandemic period, Black and Latinx beneficiaries were less likely than Whites to be hospitalized in the surge period despite minorities' heavier COVID-19 disease burden. The emergence of racial and ethnic differences in access to scarce life-preserving healthcare may hold lessons for future crises.

Antipsychotics and the risk of diabetes and death among adults with serious mental illnesses.

BACKGROUND: Individuals with schizophrenia exposed to second-generation antipsychotics (SGA) have an increased risk for diabetes, with aripiprazole purportedly a safer drug. Less is known about the drugs' mortality risk or whether serious mental illness (SMI) diagnosis or race/ethnicity modify these effects. METHODS: Authors created a retrospective cohort of non-elderly adults with SMI initiating monotherapy with an SGA (olanzapine, quetiapine, risperidone, and ziprasidone, aripiprazole) or haloperidol during 2008-2013. Three-year diabetes incidence or all-cause death risk differences were estimated between each drug and aripiprazole, the comparator, as well as effects within SMI diagnosis and race/ethnicity. Sensitivity analyses evaluated potential confounding by indication. RESULTS: 38 762 adults, 65% White and 55% with schizophrenia, initiated monotherapy, with haloperidol least (6%) and quetiapine most (26·5%) frequent. Three-year mortality was 5% and diabetes incidence 9.3%. Compared with aripiprazole, haloperidol and olanzapine reduced diabetes risk by 1.9 (95% CI 1.2-2.6) percentage points, or a 18.6 percentage point reduction relative to aripiprazole users' unadjusted risk (10.2%), with risperidone having a smaller advantage. Relative to aripiprazole users' unadjusted risk (3.4%), all antipsychotics increased mortality risk by 1.1-2.2 percentage points, representing 32.4-64.7 percentage point increases. Findings within diagnosis and race/ethnicity were generally consistent with overall findings. Only quetiapine's higher mortality risk held in sensitivity analyses. CONCLUSIONS: Haloperidol's, olanzapine's, and risperidone's lower diabetes risks relative to aripiprazole were not robust in sensitivity analyses but quetiapine's higher mortality risk proved robust. Findings expand the evidence on antipsychotics' risks, suggesting a need for caution in the use of quetiapine among individuals with SMI.

Identifying Medicare Beneficiaries With Delirium.

BACKGROUND: Each year, thousands of older adults develop delirium, a serious, preventable condition. At present, there is no well-validated method to identify patients with delirium when using Medicare claims data or other large datasets. We developed and assessed the performance of classification algorithms based on longitudinal Medicare administrative data that included International Classification of Diseases, 10th Edition diagnostic codes. METHODS: Using a linked electronic health record (EHR)-Medicare claims dataset, 2 neurologists and 2 psychiatrists performed a standardized review of EHR records between 2016 and 2018 for a stratified random sample of 1002 patients among 40,690 eligible subjects. Reviewers adjudicated delirium status (reference standard) during this 3-year window using a structured protocol. We calculated the probability that each patient had delirium as a function of classification algorithms based on longitudinal Medicare claims data. We compared the performance of various algorithms against the reference standard, computing calibration-in-the-large, calibration slope, and the area-under-receiver-operating-curve using 10-fold cross-validation (CV). RESULTS: Beneficiaries had a mean age of 75 years, were predominately female (59%), and non-Hispanic Whites (93%); a review of the EHR indicated that 6% of patients had delirium during the 3 years. Although several classification algorithms performed well, a relatively simple model containing counts of delirium-related diagnoses combined with patient age, dementia status, and receipt of antipsychotic medications had the best overall performance [CV- calibration-in-the-large <0.001, CV-slope 0.94, and CV-area under the receiver operating characteristic curve (0.88 95% confidence interval: 0.84-0.91)]. CONCLUSIONS: A delirium classification model using Medicare administrative data and International Classification of Diseases, 10th Edition diagnosis codes can identify beneficiaries with delirium in large datasets.

Significance and Factors Associated with Antipsychotic Polypharmacy Utilization Among Publicly Insured US Adults.

Antipsychotic polypharmacy (APP) lacks evidence of effectiveness in the care of schizophrenia or other disorders for which antipsychotic drugs are indicated, also exposing patients to more risks. Authors assessed APP prevalence and APP association with beneficiary race/ethnicity and payer among publicly-insured adults regardless of diagnosis. Retrospective repeated panel study of fee-for-service (FFS) Medicare, Medicaid, and dually-eligible white, black, and Latino adults residing in California, Georgia, Iowa, Mississippi, Oklahoma, South Dakota, or West Virginia, filling antipsychotic prescriptions between July 2008 and June 2013. Primary outcome was any monthly APP utilization. Across states and payers, 11% to 21% of 397,533 antipsychotic users and 12% to 19% of 9,396,741 person-months had some APP utilization. Less than 50% of person-months had a schizophrenia diagnosis and up to 19% had no diagnosed mental illness. Payer modified race/ethnicity effects on APP utilization only in CA; however, the odds of APP utilization remained lower for minorities than for whites. Elsewhere, the odds varied by race/ethnicity only in OK, with Latinos having lower odds than whites (odds ratio 0.76; 95% confidence interval 0.60-0.96). The odds of APP utilization varied by payer in several study states, with odds generally higher for Dual eligibles, although the differences were generally small; the odds also varied by year (lower at study end). APP was frequently utilized but mostly declined over time. APP utilization patterns varied across states, with no consistent association with race/ethnicity and small payer effects. Greater use of APP-reducing strategies are needed, particularly among non-schizophrenia populations.

Quality of primary health care in China: challenges and recommendations.

China has substantially increased financial investment and introduced favourable policies for strengthening its primary health care system with core responsibilities in preventing and managing chronic diseases such as hypertension and emerging infectious diseases such as coronavirus disease 2019 (COVID-19). However, widespread gaps in the quality of primary health care still exist. In this Review, we aim to identify the causes for this poor quality, and provide policy recommendations. System challenges include: the suboptimal education and training of primary health-care practitioners, a fee-for-service payment system that incentivises testing and treatments over prevention, fragmentation of clinical care and public health service, and insufficient continuity of care throughout the entire health-care system. The following recommendations merit consideration: (1) enhancement of the quality of training for primary health-care physicians, (2) establishment of performance accountability to incentivise high-quality and high-value care; (3) integration of clinical care with the basic public health services, and (4) strengthening of the coordination between primary health-care institutions and hospitals. Additionally, China should consider modernising its primary health-care system through the establishment of a learning health system built on digital data and innovative technologies.

A maximum likelihood approach to power calculations for stepped wedge designs of binary outcomes.

In stepped wedge designs (SWD), clusters are randomized to the time period during which new patients will receive the intervention under study in a sequential rollout over time. By the study's end, patients at all clusters receive the intervention, eliminating ethical concerns related to withholding potentially efficacious treatments. This is a practical option in many large-scale public health implementation settings. Little statistical theory for these designs exists for binary outcomes. To address this, we utilized a maximum likelihood approach and developed numerical methods to determine the asymptotic power of the SWD for binary outcomes. We studied how the power of a SWD for detecting risk differences varies as a function of the number of clusters, cluster size, the baseline risk, the intervention effect, the intra-cluster correlation coefficient, and the time effect. We studied the robustness of power to the assumed form of the distribution of the cluster random effects, as well as how power is affected by variable cluster size. % SWD power is sensitive to neither, in contrast to the parallel cluster randomized design which is highly sensitive to variable cluster size. We also found that the approximate weighted least square approach of Hussey and Hughes (2007, Design and analysis of stepped wedge cluster randomized trials. Contemporary Clinical Trials 28, 182-191) for binary outcomes under-estimates the power in some regions of the parameter spaces, and over-estimates it in others. The new method was applied to the design of a large-scale intervention program on post-partum intra-uterine device insertion services for preventing unintended pregnancy in the first 1.5 years following childbirth in Tanzania, where it was found that the previously available method under-estimated the power.

A Bayesian hierarchical model for characterizing the diffusion of new antipsychotic drugs.

New prescription medications are a primary driver of spending growth in the United States. For patients with severe mental illnesses, second-generation antipsychotic (SGA) medications feature prominently. However, many SGAs are costly, particularly before generic entry, and some may increase the risk of diabetes. Because physicians play a prominent role in new prescription adoption, understanding their prescribing behaviors is policy-relevant. Several features of prescription data, such as different antipsychotic choice sets over time, variable physician prescription volumes, and correlation among drug choices within physicians, complicate inferences. We propose a multivariate Bayesian hierarchical model with piecewise random effects to characterize the diffusion of new antipsychotic drugs. This model captures the complex prescriber-specific relationships among the different diffusion processes and takes advantage of the Bayesian paradigm to quantify uncertainty for all parameters straightforwardly. To evaluate the prescribing patterns for each physician, we propose various indices to identify early new SGA adopters. A sample of nearly 17,000 US physicians whose antipsychotic drug prescribing information was collected between January 1, 1997 and December 31, 2007 illustrates the methods. Determinants of high prescription rates and adoption speeds of new SGAs included physician sex, age, hospital affiliation, physician specialty, and office location. Large within- and between-provider variations in prescribing patterns of new SGAs were identified. Early adopters for one drug were not early adopters for another drug.

Hospital-Readmission Risk - Isolating Hospital Effects from Patient Effects.

BACKGROUND: To isolate hospital effects on risk-standardized hospital-readmission rates, we examined readmission outcomes among patients who had multiple admissions for a similar diagnosis at more than one hospital within a given year. METHODS: We divided the Centers for Medicare and Medicaid Services hospital-wide readmission measure cohort from July 2014 through June 2015 into two random samples. All the patients in the cohort were Medicare recipients who were at least 65 years of age. We used the first sample to calculate the risk-standardized readmission rate within 30 days for each hospital, and we classified hospitals into performance quartiles, with a lower readmission rate indicating better performance (performance-classification sample). The study sample (identified from the second sample) included patients who had two admissions for similar diagnoses at different hospitals that occurred more than 1 month and less than 1 year apart, and we compared the observed readmission rates among patients who had been admitted to hospitals in different performance quartiles. RESULTS: In the performance-classification sample, the median risk-standardized readmission rate was 15.5% (interquartile range, 15.3 to 15.8). The study sample included 37,508 patients who had two admissions for similar diagnoses at a total of 4272 different hospitals. The observed readmission rate was consistently higher among patients admitted to hospitals in a worse-performing quartile than among those admitted to hospitals in a better-performing quartile, but the only significant difference was observed when the patients were admitted to hospitals in which one was in the best-performing quartile and the other was in the worst-performing quartile (absolute difference in readmission rate, 2.0 percentage points; 95% confidence interval, 0.4 to 3.5; P=0.001). CONCLUSIONS: When the same patients were admitted with similar diagnoses to hospitals in the best-performing quartile as compared with the worst-performing quartile of hospital readmission performance, there was a significant difference in rates of readmission within 30 days. The findings suggest that hospital quality contributes in part to readmission rates independent of factors involving patients. (Funded by Yale-New Haven Hospital Center for Outcomes Research and Evaluation and others.).

Registry-Based Prospective, Active Surveillance of Medical-Device Safety.

BACKGROUND: The process of assuring the safety of medical devices is constrained by reliance on voluntary reporting of adverse events. We evaluated a strategy of prospective, active surveillance of a national clinical registry to monitor the safety of an implantable vascular-closure device that had a suspected association with increased adverse events after percutaneous coronary intervention (PCI). METHODS: We used an integrated clinical-data surveillance system to conduct a prospective, propensity-matched analysis of the safety of the Mynx vascular-closure device, as compared with alternative approved vascular-closure devices, with data from the CathPCI Registry of the National Cardiovascular Data Registry. The primary outcome was any vascular complication, which was a composite of access-site bleeding, access-site hematoma, retroperitoneal bleeding, or any vascular complication requiring intervention. Secondary safety end points were access-site bleeding requiring treatment and postprocedural blood transfusion. RESULTS: We analyzed data from 73,124 patients who had received Mynx devices after PCI procedures with femoral access from January 1, 2011, to September 30, 2013. The Mynx device was associated with a significantly greater risk of any vascular complication than were alternative vascular-closure devices (absolute risk, 1.2% vs. 0.8%; relative risk, 1.59; 95% confidence interval [CI], 1.42 to 1.78; P<0.001); there was also a significantly greater risk of access-site bleeding (absolute risk, 0.4% vs. 0.3%; relative risk, 1.34; 95% CI, 1.10 to 1.62; P=0.001) and transfusion (absolute risk, 1.8% vs. 1.5%; relative risk, 1.23; 95% CI, 1.13 to 1.34; P<0.001). The initial alerts occurred within the first 12 months of monitoring. Relative risks were greater in three prespecified high-risk subgroups: patients with diabetes, those 70 years of age or older, and women. All safety alerts were confirmed in an independent sample of 48,992 patients from April 1, 2014, to September 30, 2015. CONCLUSIONS: A strategy of prospective, active surveillance of a clinical registry rapidly identified potential safety signals among recipients of an implantable vascular-closure device, with initial alerts occurring within the first 12 months of monitoring. (Funded by the Food and Drug Administration and others.).

Medication Utilization for Alcohol Use Disorder in a Commercially Insured Population.

OBJECTIVE: Examine patterns of alcohol use disorder (AUD) medication use and identify factors associated with prescription fill among commercially insured individuals with an index AUD visit. DESIGN: Using 2008-2018 claims data from a large national insurer, estimate days to first AUD medication using cause-specific hazards approach to account for competing risk of benefits loss. PARTICIPANTS: Aged 17-64 with ≥ 1 AUD visit. MAIN MEASURE: Days to AUD medication fill. KEY RESULTS: A total of 13.3% of the 151,128 with an index visit filled an AUD prescription after that visit, while 69.8% lost benefits before filling and 17.0% remained enrolled but did not fill (median days observed = 305). Almost half (46.3%) of those who filled a prescription received substance use disorder (SUD) inpatient care within 7 days before the fill, and 63.4% received SUD outpatient care. Likelihood of medication use was higher for those aged 26-35, 36-45, and 46-55 years relative to 56-64 years (e.g., 26-35: hazard ratio = 1.29 [95% confidence interval 1.23-1.36]); those diagnosed with moderate/severe AUD (2.05 [1.98-2.12]), co-occurring opioid use disorder (OUD) (1.33 [1.26-1.39]), or severe mental illness (1.31 [1.27-1.35]); those with a chronic alcohol-related diagnosis (1.08 [1.04-1.12]); and those whose index visit was in an inpatient/emergency department (1.27 [1.23-1.31]) or intermediate care setting (1.13 [1.07-1.20]) relative to outpatient. Likelihood of use was higher in later years relative to 2008 (e.g., 2018:2.02 [1.89-2.15]) and higher for those who received the majority of AUD care in a practice with a psychiatrist/addiction medicine specialist (1.13 [1.10-1.16]). Likelihood of use was lower for those diagnosed with a SUD other than AUD or OUD (0.88 [0.85-0.92]), those with an acute alcohol-related condition (0.79 [0.75-0.84]), and males (0.71 [0.69-0.73]). CONCLUSIONS: While AUD medication use increased and was more common among individuals with greater severity, few patients who could benefit from medications are using them. More efforts are needed to identify and treat individuals in non-acute care settings earlier in their course of AUD.

Life Expectancy after Myocardial Infarction, According to Hospital Performance.

BACKGROUND: Thirty-day risk-standardized mortality rates after acute myocardial infarction are commonly used to evaluate and compare hospital performance. However, it is not known whether differences among hospitals in the early survival of patients with acute myocardial infarction are associated with differences in long-term survival. METHODS: We analyzed data from the Cooperative Cardiovascular Project, a study of Medicare beneficiaries who were hospitalized for acute myocardial infarction between 1994 and 1996 and who had 17 years of follow-up. We grouped hospitals into five strata that were based on case-mix severity. Within each case-mix stratum, we compared life expectancy among patients admitted to high-performing hospitals with life expectancy among patients admitted to low-performing hospitals. Hospital performance was defined by quintiles of 30-day risk-standardized mortality rates. Cox proportional-hazards models were used to calculate life expectancy. RESULTS: The study sample included 119,735 patients with acute myocardial infarction who were admitted to 1824 hospitals. Within each case-mix stratum, survival curves of the patients admitted to hospitals in each risk-standardized mortality rate quintile separated within the first 30 days and then remained parallel over 17 years of follow-up. Estimated life expectancy declined as hospital risk-standardized mortality rate quintile increased. On average, patients treated at high-performing hospitals lived between 0.74 and 1.14 years longer, depending on hospital case mix, than patients treated at low-performing hospitals. When 30-day survivors were examined separately, there was no significant difference in unadjusted or adjusted life expectancy across hospital risk-standardized mortality rate quintiles. CONCLUSIONS: In this study, patients admitted to high-performing hospitals after acute myocardial infarction had longer life expectancies than patients treated in low-performing hospitals. This survival benefit occurred in the first 30 days and persisted over the long term. (Funded by the National Heart, Lung, and Blood Institute and the National Institute of General Medical Sciences Medical Scientist Training Program.).

Claims-based cardiovascular outcome identification for clinical research: Results from 7 large randomized cardiovascular clinical trials.

BACKGROUND: Medicare insurance claims may provide an efficient means to ascertain follow-up of older participants in clinical research. We sought to determine the accuracy and completeness of claims- versus site-based follow-up with clinical event committee (+CEC) adjudication of cardiovascular outcomes. METHODS: We performed a retrospective study using linked Medicare and Duke Database of Clinical Trials data. Medicare claims were linked to clinical data from 7 randomized cardiovascular clinical trials. Of 52,476 trial participants, linking resulted in 5,839 (of 10,497 linkage-eligible) Medicare-linked trial participants with fee-for-service A and B coverage. Death, myocardial infarction (MI), stroke, and revascularization incidences were compared using Medicare inpatient claims only, site-reported events (+CEC) only, or a combination of the 2. Randomized treatment effects were compared as a function of whether claims-based, site-based (+CEC), or a combined system was used for event detection. RESULTS: Among the 5,839 study participants, the annual event rates were similar between claims- and site-based (+CEC) follow-up: death (overall rate 5.2% vs 5.2%; adjusted κ 0.99), MI (2.2% vs 2.3%; adjusted κ 0.96), stroke (0.7% vs 0.7%; adjusted κ 0.99), and any revascularization (7.4% vs 7.9%; adjusted κ 0.95). Of events detected by claims yet not reported by CEC, a minority were reported by sites but negatively adjudicated by CEC (39% of MIs and 18% of strokes). Differences in individual case concordance led to higher event rates when claims- and site-based (+CEC) systems were combined. Randomized treatment effects were similar among the 3 approaches for each outcome of interest. CONCLUSIONS: Claims- versus site-based (+CEC) follow-up identified similar overall cardiovascular event rates despite meaningful differences in the events detected. Randomized treatment effects were similar using the 2 methods, suggesting claims data could be used to support clinical research leveraging routinely collected data. This approach may lead to more effective evidence generation, synthesis, and appraisal of medical products and inform the strategic approaches toward the National Evaluation System for Health Technology.

Double robust estimation for multiple unordered treatments and clustered observations: Evaluating drug-eluting coronary artery stents.

Postmarket comparative effectiveness and safety analyses of therapeutic treatments typically involve large observational cohorts. We propose double robust machine learning estimation techniques for implantable medical device evaluations where there are more than two unordered treatments and patients are clustered in hospitals. This flexible approach also accommodates high-dimensional covariates drawn from clinical databases. The Massachusetts Data Analysis Center percutaneous coronary intervention cohort is used to assess the composite outcome of 10 drug-eluting stents among adults implanted with at least one drug-eluting stent in Massachusetts. We find remarkable discrimination between stents. A simulation study designed to mimic this coronary intervention cohort is also presented and produced similar results.

Approaches to treatment effect heterogeneity in the presence of confounding.

The literature on causal effect estimation tends to focus on the population mean estimand, which is less informative as medical treatments are becoming more personalized and there is increasing awareness that subpopulations of individuals may experience a group-specific effect that differs from the population average. In fact, it is possible that there is underlying systematic effect heterogeneity that is obscured by focusing on the population mean estimand. In this context, understanding which covariates contribute to this treatment effect heterogeneity (TEH) and how these covariates determine the differential treatment effect (TE) is an important consideration. Towards such an understanding, this paper briefly reviews three approaches used in making causal inferences and conducts a simulation study to compare these approaches according to their performance in an exploratory evaluation of TEH when the heterogeneous subgroups are not known a priori. Performance metrics include the detection of any heterogeneity, the identification and characterization of heterogeneous subgroups, and unconfounded estimation of the TE within subgroups. The methods are then deployed in a comparative effectiveness evaluation of drug-eluting versus bare-metal stents among 54 099 Medicare beneficiaries in the continental United States admitted to a hospital with acute myocardial infarction in 2008.

The relation of telomere length at midlife to subsequent 20-year depression trajectories among women.

BACKGROUND: Telomeres cap and protect DNA but shorten with each somatic cell division. Aging and environmental and lifestyle factors contribute to the speed of telomere attrition. Current evidence suggests a link between relative telomere length (RTL) and depression but the directionality of the relationship remains unclear. We prospectively examined associations between RTL and subsequent depressive symptom trajectories. METHODS: Among 8,801 women of the Nurses' Health Study, depressive symptoms were measured every 4 years from 1992 to 2012; group-based trajectories of symptoms were identified using latent class growth-curve analysis. Multinomial logistic models were used to relate midlife RTLs to the probabilities of assignment to subsequent depressive symptom trajectory groups. RESULTS: We identified four depressive symptom trajectory groups: minimal depressive symptoms (62%), worsening depressive symptoms (14%), improving depressive symptoms (19%), and persistent-severe depressive symptoms (5%). Longer midlife RTLs were related to significantly lower odds of being in the worsening symptoms trajectory versus minimal trajectory but not to other trajectories. In comparison with being in the minimal symptoms group, the multivariable-adjusted odds ratio of being in the worsening depressive symptoms group was 0.78 (95% confidence interval, 0.62-0.97; p = 0.02), for every standard deviation increase in baseline RTL. CONCLUSIONS: In this large prospective study of generally healthy women, longer telomeres at midlife were associated with significantly lower risk of a subsequent trajectory of worsening mood symptoms over 20 years. The results raise the possibility of telomere shortening as a novel contributing factor to late-life depression.

Trends in the use of coercive measures in Finnish psychiatric hospitals: a register analysis of the past two decades.

BACKGROUND: Coercive measures is a topic that has long been discussed in the field of psychiatry. Despite global reports of reductions in the use of restraint episodes due to new regulations, it is still questionable if practices have really changed over time. For this study, we examined the rates of coercive measures in the inpatient population of psychiatric care providers across Finland to identify changing trends as well as variations in such trends by region. METHODS: In this nationwide registry analysis, we extracted patient data from the national database (The Finnish National Care Register for Health Care) over a 20-year period. We included adult patients admitted to psychiatric units (care providers) and focused on patients who had faced coercive measures (seclusion, limb restraints, forced injection and physical restraints) during their hospital stay. Multilevel logistical models (a polynomial model of quadratic form) were used to examine trends in prevalence of any coercive measures as well as the other four specified coercive measures over time, and to investigate variation in such trends among care providers and regions. RESULTS: Between 1995 and 2014, the dataset contained 226,948 inpatients who had been admitted during the 20-year time frame (505,169 treatment periods). The overall prevalence of coercive treatment on inpatients was 9.8%, with a small decrease during 2011-2014. The overall prevalence of seclusion, limb restraints, forced injection and physical restraints on inpatients was 6.9, 3.8, 2.6 and 0.8%, respectively. Only the use of limb restraints showed a downward trend over time. Geographic and care provider variations in specific coercive measures used were also observed. CONCLUSIONS: Despite the decreasing national level of coercive measures used in Finnish psychiatric hospitals, the overall reduction has been small during the last two decades. These results have implications on the future development of structured guidelines and interventions for preventing and more effectively managing challenging situations. Clinical guidelines and staff education related to the use of coercive measures should be critically assessed to ensure that the staff members working with vulnerable patient populations in psychiatric hospitals are ethically competent.