Assessing the properties of patient-specific treatment effect estimates from causal forest algorithms under essential heterogeneity.

BACKGROUND: Treatment variation from observational data has been used to estimate patient-specific treatment effects. Causal Forest Algorithms (CFAs) developed for this task have unknown properties when treatment effect heterogeneity from unmeasured patient factors influences treatment choice - essential heterogeneity. METHODS: We simulated eleven populations with identical treatment effect distributions based on patient factors. The populations varied in the extent that treatment effect heterogeneity influenced treatment choice. We used the generalized random forest application (CFA-GRF) to estimate patient-specific treatment effects for each population. Average differences between true and estimated effects for patient subsets were evaluated. RESULTS: CFA-GRF performed well across the population when treatment effect heterogeneity did not influence treatment choice. Under essential heterogeneity, however, CFA-GRF yielded treatment effect estimates that reflected true treatment effects only for treated patients and were on average greater than true treatment effects for untreated patients. CONCLUSIONS: Patient-specific estimates produced by CFAs are sensitive to why patients in real-world practice make different treatment choices. Researchers using CFAs should develop conceptual frameworks of treatment choice prior to estimation to guide estimate interpretation ex post.

ICD-10 diagnosis codes in electronic health records do not adequately capture fracture complexity for proximal humerus fractures.

BACKGROUND: The ability to do comparative effectiveness research (CER) for proximal humerus fractures (PHF) using data in electronic health record (EHR) systems and administrative claims databases was enhanced by the 10th revision of the International Classification of Diseases (ICD-10), which expanded the diagnosis codes for PHF to describe fracture complexity including displacement and the number of fracture parts. However, these expanded codes only enhance secondary use of data for research if the codes selected and recorded correctly reflect the fracture complexity. The objective of this project was to assess the accuracy of ICD-10 diagnosis codes documented during routine clinical practice for secondary use of EHR data. METHODS: A sample of patients with PHFs treated by orthopedic providers across a large, regional health care system between January 1, 2016, and December 31, 2018, were retrospectively identified from the EHR. Four fellowship-trained orthopedic surgeons reviewed patient radiographs and recorded the Neer Classification characteristics of displacement, number of parts, and fracture location(s). The fracture characteristics were then reviewed by a trained coder, and the most clinically appropriate ICD-10 diagnosis code based on the number of fracture parts was assigned. We assessed congruence between ICD-10 codes documented in the EHR and radiograph-validated codes, and assessed sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for EHR-documented ICD-10 codes. RESULTS: There were 761 patients with unilateral, closed PHF who met study inclusion criteria. On average, patients were 67 years of age and 77% were female. Based on radiograph review, 37% were 1-part fractures, 42% were 2-part, 11% were 3-part, and 10% were 4-part fractures. Of the EHR diagnosis codes recorded during clinical practice, 59% were "unspecified" fracture diagnosis codes that did not identify the number of fracture parts. Examination of fracture codes revealed PPV was highest for 1-part (PPV = 0.66, 95% confidence interval [CI] 0.60-0.72) and 4-part fractures (PPV = 0.67, 95% CI 0.13-1.00). CONCLUSIONS: Current diagnosis coding practices do not adequately capture the fracture complexity needed to conduct subgroup analysis for PHF. Conclusions drawn from population studies or large databases using ICD-10 codes for PHF classification should be interpreted within this limitation. Future studies are warranted to improve diagnostic coding to support large observational studies using EHR and administrative claims data.

Unanticipated costs associated with interscalene nerve catheters for shoulder surgery.

BACKGROUND: Regional anesthesia has become a mainstay of analgesia following shoulder arthroscopic and reconstructive procedures. Local anesthetic can be injected in the perineural space of the brachial plexus by a single shot or continuously by an indwelling catheter. Although previous studies have compared efficacy and direct cost of single shot to catheters, few have evaluated unanticipated costs of ongoing care or complications. Pulmonary complications can lead to unexpected admissions and emergency department visits. The purpose of the study was to identify unplanned hospital admissions or emergency department visits related to regional anesthesia after shoulder surgery and determine the additional associated costs. METHODS: A series of 1888 shoulder surgeries were identified in 1856 unique patients at a single, large academic center. As part of an interscalene nerve catheter program, a continuous interscalene block (CIB) was given to 1728 patients, whereas 160 patients had a single-shot interscalene block (SSIB). A hospital-employed quality control nurse contacted all patients receiving a CIB at 1, 2, 7, and 14 days following surgery. All emergency department visits and readmissions were recorded, and the associated billing charges were reviewed for the inpatient and any outpatient visits immediately preceding or immediately following the readmission. The regional average Medicare fee schedule was used to determine a cost for these episodes of care. RESULTS: Of the 1728 patients who had CIB, 10 patients were readmitted following open or arthroscopic surgery or presented to the emergency department in the immediate postoperative period for pulmonary compromise. No patient in the SSIB group had an emergency department visit or readmission. The average age of the 10 patients with readmission was 60 years (7 females, 3 males). The majority were diagnosed with hypoxemia on admission (R09.02). Length of stay during readmission ranged from 0 to 4 days, with 1 patient requiring admission to the intensive care unit. The average cost of admission to the hospital or visit to the emergency department was $6849 (range, $1988-$19,483). These costs were primarily related to chest radiographs and electrocardiogram (9/10), chest computed tomography (CT) with contrast (3/10), and head CT (2/10). CONCLUSION: Although uncommon, unanticipated pulmonary complications after CIB can result in significant cost compared to SSIB. The indirect costs of pulmonary workup after readmission or emergency department workup may be overlooked if only considering direct costs, such as medication charges, medical supplies, and physician fees.

Assessing the ability of an instrumental variable causal forest algorithm to personalize treatment evidence using observational data: the case of early surgery for shoulder fracture.

BACKGROUND: Comparative effectiveness research (CER) using observational databases has been suggested to obtain personalized evidence of treatment effectiveness. Inferential difficulties remain using traditional CER approaches especially related to designating patients to reference classes a priori. A novel Instrumental Variable Causal Forest Algorithm (IV-CFA) has the potential to provide personalized evidence using observational data without designating reference classes a priori, but the consistency of the evidence when varying key algorithm parameters remains unclear. We investigated the consistency of IV-CFA estimates through application to a database of Medicare beneficiaries with proximal humerus fractures (PHFs) that previously revealed heterogeneity in the effects of early surgery using instrumental variable estimators. METHODS: IV-CFA was used to estimate patient-specific early surgery effects on both beneficial and detrimental outcomes using different combinations of algorithm parameters and estimate variation was assessed for a population of 72,751 fee-for-service Medicare beneficiaries with PHFs in 2011. Classification and regression trees (CART) were applied to these estimates to create ex-post reference classes and the consistency of these classes were assessed. Two-stage least squares (2SLS) estimators were applied to representative ex-post reference classes to scrutinize the estimates relative to known 2SLS properties. RESULTS: IV-CFA uncovered substantial early surgery effect heterogeneity across PHF patients, but estimates for individual patients varied with algorithm parameters. CART applied to these estimates revealed ex-post reference classes consistent across algorithm parameters. 2SLS estimates showed that ex-post reference classes containing older, frailer patients with more comorbidities, and lower utilizers of healthcare were less likely to benefit and more likely to have detriments from higher rates of early surgery. CONCLUSIONS: IV-CFA provides an illuminating method to uncover ex-post reference classes of patients based on treatment effects using observational data with a strong instrumental variable. Interpretation of treatment effect estimates within each ex-post reference class using traditional CER methods remains conditional on the extent of measured information in the data.

Factors related to initial treatment for adhesive capsulitis in the medicare population.

BACKGROUND: Primary adhesive capsulitis (AC) is not well understood, and controversy remains about the most effective treatment approaches. Even less is known about the treatment of AC in the Medicare population. We aimed to fully characterize initial treatment for AC in terms of initial treatment utilization, timing of initial treatments and treatment combinations. METHODS: Using United States Medicare claims from 2010-2012, we explored treatment utilization and patient characteristics associated with initial treatment for primary AC among 7,181 Medicare beneficiaries. Patients with primary AC were identified as patients seeking care for a new shoulder complaint in 2011, with the first visit related to shoulder referred to as the index date, an x-ray or MRI of the shoulder region, and two separate diagnoses of AC (ICD-9-CM codes: 726.00). The treatment period was defined as the 90 days immediately following the index shoulder visit. A multivariable logistic model was used to assess baseline patient factors associated with receiving surgery within the treatment period. RESULTS: Ninety percent of beneficiaries with primary AC received treatment within 90 days of their index shoulder visit. Physical therapy (PT) alone (41%) and injection combined with PT (34%) were the most common treatment approaches. Similar patient profiles emerged across treatment groups, with higher proportions of racial minorities, socioeconomically disadvantaged and more frail patients favoring injections or watchful waiting. Black beneficiaries (OR = 0.37, [0.16, 0.86]) and those residing in the northeast (OR = 0.36, [0.18, 0.69]) had significantly lower odds of receiving surgery in the treatment period. Conversely, younger beneficiaries aged 66-69 years (OR = 6.75, [2.12, 21.52]) and 70-75 years (OR = 5.37, [1.67, 17.17]) and beneficiaries with type 2 diabetes had significantly higher odds of receiving surgery (OR = 1.41, [1.03, 1.92]). CONCLUSIONS: Factors such as patient baseline health and socioeconomic characteristics appear to be important for physicians and Medicare beneficiaries making treatment decisions for primary AC.

Regenesorb and polylactic acid hydroxyapatite anchors are associated with similar osseous integration and rotator cuff healing at 2 years.

BACKGROUND: Commercially available suture anchors for rotator cuff repairs can differ significantly in architecture and material. Clinical data on their osseous integration and its effect on patient-reported outcomes is scarce. Preclinical investigations indicated a higher rate of osseous integration for the open-architecture design of the Healicoil Regenesorb anchor than the closed-threaded design of the Twinfix (Smith & Nephew). The purpose of this study was to investigate these 2 anchors with different architecture and material to determine their effect on osseous integration and clinical outcomes after rotator cuff repair. METHODS: A prospective randomized controlled trial was performed from 2014 to 2019. Sixty-four patients (39 females, 25 males) with an average age of 58.7 years who underwent arthroscopic rotator cuff repair by one of 4 board-certified, fellowship-trained surgeons were randomized to receive Healicoil Regenesorb (PLGA/ß-TCP/Calcium Sulfate) or Twinfix Ultra HA (PLLA/HA) anchors. Thirty-two patients had Healicoil anchors implanted, and 32 patients had Twinfix anchors implanted. Of the 64 patients, 51 returned at 24 months for computed tomographic (CT) examination (25 Twinfix and 26 Healicoil) to determine osteointegration of the anchors. Patient-reported outcomes, including Penn Shoulder Score (PENN), Western Ontario Rotator Cuff Index, visual analog scale, EQ-5D, Single Assessment Numeric Evaluation, Global Rating of Change, were collected at baseline, 6 weeks, 3 months, 6 months, 12 months, and 24 months. Ultrasonography was used to assess rotator cuff integrity after 6 months. Two board-certified, fellowship-trained orthopedic surgeons, blinded to the type of anchors, analyzed the CT scans to assess the anchor osteointegration at 24 months using a previously published grading scale. RESULTS: There were no differences in demographics, preoperative outcomes, or baseline characteristics such as tear size, number of anchors, Goutallier classification, or smoking status between groups. There was no difference in osseous integration between the 2 anchors at 24 months (P = .117). Eight patients had rotator cuff retears, of which 2 patients had Twinfix anchors and 6 patients had Healicoil anchors (P = .18). There were no statistically significant differences in patient-reported outcomes or complications between groups. The 2-year PENN scores were 89 with the Twinfix and 88 with Healicoil anchors (P = .55). CONCLUSION: Despite differences in material and anchor architecture, the rate of healing and patient-reported outcomes were similar between the Twinfix and Healicoil anchor groups. The rate of osteointegration was the same at 2 years.

Freestanding Dialysis Facility Quality Incentive Program Scores and Mortality Among Incident Dialysis Patients in the United States.

RATIONALE & OBJECTIVE: The Centers for Medicare & Medicaid Services introduced the Quality Incentive Program (QIP) along with the bundled payment reform to improve the quality of dialysis care in the United States. The QIP has been criticized for using easily obtained laboratory indicators without patient-centered measures and for a lack of evidence for an association between QIP indicators and patient outcomes. This study examined the association between dialysis facility QIP performance scores and survival among patients after initiation of dialysis. STUDY DESIGN: Retrospective cohort study. SETTING & PARTICIPANTS: Study participants included 84,493 patients represented in the US Renal Disease System's patient-level data who had initiated dialysis between January 1, 2013, and December 1, 2013, and who did not, during the first 90 days after dialysis initiation, die, receive a transplant, or become lost to follow-up. Patients were followed up for the study outcome through March 31, 2014. PREDICTOR: Dialysis facility QIP scores. OUTCOME: Mortality. ANALYTICAL APPROACH: Using a unique facility identifier, we linked Medicare freestanding dialysis facility data from 2015 with US Renal Disease System patient-level data. Kaplan-Meier product limit estimator was used to describe the survival of study participants. Cox proportional hazards regression was used to assess the multivariable association between facility performance scores and patient survival. RESULTS: Excluding patients who died during the first 90 days of dialysis, 11.8% of patients died during an average follow-up of 5 months. Facilities with QIP scores<45 (HR, 1.39; 95% CI, 1.15-1.68) and 45 to<60 (HR, 1.21; 95% CI, 1.10-1.33) had higher patient mortality rates than facilities with scores≥90. LIMITATIONS: Because the Centers for Medicare & Medicaid Services have revised QIP criteria each year, the findings may not relate to years other than those studied. CONCLUSIONS: Dialysis facilities characterized by lower QIP scores were associated with higher rates of patient mortality. These findings need to be replicated to assess their consistency over time.

Geographic variation in inhaled corticosteroid use for children with persistent asthma in Medicaid.

OBJECTIVE: Geographic variation in the rates of inhaled corticosteroid (ICS) use for children with persistent asthma in Medicaid has been reported, but the source of this variation is unknown. The objective of this study was to quantify the geographic variation in ICS use for children with persistent asthma in Medicaid that remains after adjusting for the characteristics of children in an area. METHODS: Data from the 2005-2007 Medicaid Analytic eXtract files were used. Frequent fills of short-acting beta2-agonist (SABA) were used to identify children 5-18 years of age with persistent asthma across the United States. A child was considered to have used an ICS if the child initially filled an ICS following frequent SABA use. Areas were determined using published methods, and the unadjusted ICS rate and the area treatment ratio for ICS, which adjusted for demographic and clinical characteristics, were calculated for each area. RESULTS: Of 15,917 children, 13% used an ICS. The median unadjusted ICS rate for all areas was 10% but ranged from 0% to 64%. ICS use was less than expected for more than half of the areas based on the characteristics of the children in the area, but use was nearly five times what was expected in some areas. Areas with higher than expected ICS use were found contiguous to areas with lower than expected use. CONCLUSIONS: Geographic variation in ICS not attributable to the demographic and clinical characteristics of the children in an area exists and could prove useful in the struggle to reduce asthma exacerbation rates.

A comparison of one-year treatment utilization for shoulder osteoarthritis patients initiating care with non-orthopaedic physicians and orthopaedic specialists.

BACKGROUND: In this paper we investigate patients seeking care for a new diagnosis of shoulder osteoarthritis (OA) and the association between a patient's initial physician specialty choice and one-year surgical and conservative treatment utilization. METHODS: Using retrospective data from a single large regional healthcare system, we identified 572 individuals with a new diagnosis of shoulder OA and identified the specialty of the physician which was listed as the performing physician on the index shoulder visit. We assessed treatment utilization in the year following the index shoulder visit for patients initiating care with a non-orthopaedic physician (NOP) or an orthopaedic specialist (OS). Descriptive statistics were calculated for each group and subsequent one-year surgical and conservative treatment utilization was compared between groups. RESULTS: Of the 572 patients included in the study, 474 (83%) received care from an OS on the date of their index shoulder visit, while 98 (17%) received care from a NOP. There were no differences in baseline patient age, gender, BMI or pain scores between groups. OS patients reported longer symptom duration and a higher rate of comorbid shoulder diagnoses. Patients initiating care with an OS on average received their first treatment much faster than patients initiating care with NOP (16.3 days [95% CI, 12.8, 19.7] vs. 32.3 days [95% CI, 21.0, 43.6], Z = 4.9, p < 0.01). Additionally, patients initiating care with an OS had higher odds of receiving surgery (OR = 2.65, 95% CI: 1.42, 4.95) in the year following their index shoulder visit. CONCLUSIONS: Patients initiating care with an OS received treatment much faster and were treated with more invasive services over the year following their index shoulder visit. Future work should compare patient-reported outcomes across patient groups to assess whether more expensive and invasive treatments yield better outcomes for patients with shoulder OA.

Variation in geographic access to chemotherapy by definitions of providers and service locations: a population-based observational study.

BACKGROUND: An aging population, with its associated rise in cancer incidence and strain on the oncology workforce, will continue to motivate patients, healthcare providers and policy makers to better understand the existing and growing challenges of access to chemotherapy. Administrative data, and SEER-Medicare data in particular, have been used to assess patterns of healthcare utilization because of its rich information regarding patients, their treatments, and their providers. To create measures of geographic access to chemotherapy, patients and oncologists must first be identified. Others have noted that identifying chemotherapy providers from Medicare claims is not always straightforward, as providers may report multiple or incorrect specialties and/or practice in multiple locations. Although previous studies have found that specialty codes alone fail to identify all oncologists, none have assessed whether various methods of identifying chemotherapy providers and their locations affect estimates of geographic access to care. METHODS: SEER-Medicare data was used to identify patients, physicians, and chemotherapy use in this population-based observational study. We compared two measures of geographic access to chemotherapy, local area density and distance to nearest provider, across two definitions of chemotherapy provider (identified by specialty codes or billing codes) and two definitions of chemotherapy service location (where chemotherapy services were proven to be or possibly available) using descriptive statistics. Access measures were mapped for three representative registries. RESULTS: In our sample, 57.2 % of physicians who submitted chemotherapy claims reported a specialty of hematology/oncology or medical oncology. These physicians were associated with 91.0 % of the chemotherapy claims. When providers were identified through billing codes instead of specialty codes, an additional 50.0 % of beneficiaries (from 23.8 % to 35.7 %) resided in the same ZIP code as a chemotherapy provider. Beneficiaries were also 1.3 times closer to a provider, in terms of driving time. Our access measures did not differ significantly across definitions of service location. CONCLUSIONS: Measures of geographic access to care were sensitive to definitions of chemotherapy providers; far more providers were identified through billing codes than specialty codes. They were not sensitive to definitions of service locations, as providers, regardless of how they are identified, generally provided chemotherapy at each of their practice locations.

Increased statin prescribing does not lower pneumonia risk.

BACKGROUND: Investigators have attributed protective effects of statins against pneumonia and other infections. However, these reports are based on observational data where treatments are not assigned randomly. We aimed to determine if the protective effects of statins against pneumonia are due to nonrandom treatment assignment. METHODS: We built a cohort consisting of 124 695 Medicare beneficiaries diagnosed with an acute myocardial infarction (AMI) for which we had complete claims data. We considered patients who survived at least 30 days post-AMI (full sample), or who survived for 1 year post-AMI (survivors). First, we used ordinary least squares (OLS) and logit models to determine if receiving a statin was protective against pneumonia. Second, to control for nonrandom treatment assignment, we performed an instrumental variables analysis using geographic treatment rates as an instrument. All models included patient demographics, medications, diagnoses, length of hospital stay, and out-of-pocket drug costs as covariates. Our outcome measure was a pneumonia diagnosis during the 1 year following AMI. RESULTS: A total of 76 994 patients (61.9%) filled a statin prescription, and 19 078 (15.3%) were diagnosed with pneumonia. Using OLS, the statin coefficient was -0.016 (P < .001), indicating that statins are associated with a reduction in pneumonia. Using instrumental variables, we find that statin prescriptions are not associated with a reduction in pneumonia. For the full sample, statin coefficients ranged from -0.001 to -0.01 (P > .6). CONCLUSIONS: For patients with AMI, the protective effect of statins against pneumonia is most likely the result of nonrandom treatment assignment (ie, a healthy-user bias).

Statin use after acute myocardial infarction by patient complexity: are the rates right?

BACKGROUND: Guidelines suggest statin use after acute myocardial infarction (AMI) should be close to universal in patients without safety concerns yet rates are much lower than recommended, decline with patient complexity, and display substantial geographic variation. Trial exclusions have resulted in little evidence to guide statin prescribing for complex patients. OBJECTIVE: To assess the benefits and risks associated with higher rates of statin use after AMI by baseline patient complexity. RESEARCH DESIGN: Sample includes Medicare fee-for-service patients with AMIs in 2008-2009. Instrumental variable estimators using variation in local area prescribing patterns by statin intensity as instruments were used to assess the association of higher statin prescribing rates by statin intensity on 1-year survival, adverse events, and cost by patient complexity. RESULTS: Providers seem to have individualized statin use across patients based on potential risks. Higher statin rates for noncomplex AMI patients were associated with increased survival rates with little added adverse event risk. Higher statin rates for complex AMI patients were associated with tradeoffs between higher survival rates and higher rates of adverse events. CONCLUSIONS: Higher rates of statin use for noncomplex AMI patients are associated with outcome rate changes similar to existing evidence. For the complex patients in our study, who were least represented in existing trials, higher statin-use rates were associated with survival gains and higher adverse event risks not previously documented. Policy interventions promoting higher statin-use rates for complex patients may need to be reevaluated taking careful consideration of these tradeoffs.

Geographic variation in statin use for complex acute myocardial infarction patients: evidence of effective care?

BACKGROUND: Despite strong evidence to designate statin use for secondary prevention of cardiovascular disease (CVD) as "effective care," observational studies show that many patients with CVD do not receive statins. This suggests that statin prescribing decisions for complex CVD patients are preference sensitive. OBJECTIVES: The aim of this study was to evaluate local area variation in statin prescribing for subsets of complex patients after acute myocardial infarction (AMI) to assess whether current statin prescribing patterns fit profiles of either "effective care" or "preference-sensitive care." RESEARCH DESIGN AND SUBJECTS: This was a retrospective cohort study of 124,618 Medicare patients with fee-for-service parts A, B, and D benefits who were hospitalized with AMI in 2008 or 2009 with no evidence of AMI in the past 12 months. MEASURES: Patient complexity was defined by the presence of diabetes, heart failure, and chronic kidney disease in the year before AMI admission. Local area practice styles for "no statin," "lower-intensity statins," and "high-intensity statins" were measured using the driving area for clinical care method. Statin prescribing rates for complex patient subsets were contrasted across patients grouped by local areas practice styles. RESULTS: Lower statin treatment rates were observed for patients with complex conditions, especially among those with heart failure. However, substantial local area variation in statin prescribing is observed across all complex patient groups. CONCLUSIONS: Despite guidelines promoting the use of statins for secondary prevention for CVD patients, substantial local area variation suggests that patient and provider beliefs and preferences weigh heavily in statin prescribing decisions.

Effect of a care transition intervention by pharmacists: an RCT.

BACKGROUND: Pharmacists may improve medication-related outcomes during transitions of care. The aim of the Iowa Continuity of Care Study was to determine if a pharmacist case manager (PCM) providing a faxed discharge medication care plan from a tertiary care institution to primary care could improve medication appropriateness and reduce adverse events, rehospitalization and emergency department visits. METHODS: Design. Randomized, controlled trial of 945 participants assigned to enhanced, minimal and usual care groups conducted 2007 to 2012. Subjects. Participants with cardiovascular-related conditions and/or asthma or chronic obstructive pulmonary disease were recruited from the University of Iowa Hospital and Clinics following admission to general medicine, family medicine, cardiology or orthopedics. Intervention. The minimal group received admission history, medication reconciliation, patient education, discharge medication list and medication recommendations to inpatient team. The enhanced group also received a faxed medication care plan to their community physician and pharmacy and telephone call 3-5 days post-discharge. Participants were followed for 90 days post-discharge. Main Outcomes and Measures. Medication appropriateness index (MAI), adverse events, adverse drug events and post-discharge healthcare utilization were compared by study group using linear and logistic regression, as models accommodating random effects due to pharmacists indicated little clustering. RESULTS: Study groups were similar at baseline and the intervention fidelity was high. There were no statistically significant differences by study group in medication appropriateness, adverse events or adverse drug events at discharge, 30-day and 90-day post-discharge. The average MAI per medication as 0.53 at discharge and increased to 0.75 at 90 days, and this was true across all study groups. Post-discharge, about 16% of all participants experienced an adverse event, and this did not differ by study group (p > 0.05). Almost one-third of all participants had any type of healthcare utilization within 30 days post-discharge, where 15% of all participants had a 30-day readmission. Healthcare utilization post-discharge was not statistically significant different at 30 or 90 days by study group. CONCLUSION: The pharmacist case manager did not affect medication use outcomes post-discharge perhaps because quality of care measures were high in all study groups. TRIAL REGISTRATION: Clinicaltrials.gov registration: NCT00513903, August 7, 2007.

Human-centered Design of a Health Recommender System for Orthopaedic Shoulder Treatment.

Background: Rich data on diverse patients and their treatments and outcomes within Electronic Health Record (EHR) systems can be used to generate real world evidence. A health recommender system (HRS) framework can be applied to a decision support system application to generate data summaries for similar patients during the clinical encounter to assist physicians and patients in making evidence-based shared treatment decisions. Objective: A human-centered design (HCD) process was used to develop a HRS for treatment decision support in orthopaedic medicine, the Informatics Consult for Individualized Treatment (I-C-IT). We also evaluate the usability and utility of the system from the physician's perspective, focusing on elements of utility and shared decision-making in orthopaedic medicine. Methods: The HCD process for I-C-IT included 6 steps across three phases of analysis, design, and evaluation. A team of informaticians and comparative effectiveness researchers directly engaged with orthopaedic surgeon subject matter experts in a collaborative I-C-IT prototype design process. Ten orthopaedic surgeons participated in a mixed methods evaluation of the I-C-IT prototype that was produced. Results: The HCD process resulted in a prototype system, I-C-IT, with 14 data visualization elements and a set of design principles crucial for HRS for decision support. The overall standard system usability scale (SUS) score for the I-C-IT Webapp prototype was 88.75 indicating high usability. In addition, utility questions addressing shared decision-making found that 90% of orthopaedic surgeon respondents either strongly agreed or agreed that I-C-IT would help them make data informed decisions with their patients. Conclusion: The HCD process produced an HRS prototype that is capable of supporting orthopaedic surgeons and patients in their information needs during clinical encounters. Future research should focus on refining I-C-IT by incorporating patient feedback in future iterative cycles of system design and evaluation.

Patient preferences as human factors for health data recommender systems and shared decision making in orthopaedic practice.

Background: A core set of requirements for designing AI-based Health Recommender Systems (HRS) is a thorough understanding of human factors in a decision-making process. Patient preferences regarding treatment outcomes can be one important human factor. For orthopaedic medicine, limited communication may occur between a patient and a provider during the short duration of a clinical visit, limiting the opportunity for the patient to express treatment outcome preferences (TOP). This may occur despite patient preferences having a significant impact on achieving patient satisfaction, shared decision making and treatment success. Inclusion of patient preferences during patient intake and/or during the early phases of patient contact and information gathering can lead to better treatment recommendations. Aim: We aim to explore patient treatment outcome preferences as significant human factors in treatment decision making in orthopedics. The goal of this research is to design, build, and test an app that collects baseline TOPs across orthopaedic outcomes and reports this information to providers during a clinical visit. This data may also be used to inform the design of HRSs for orthopaedic treatment decision making. Methods: We created a mobile app to collect TOPs using a direct weighting (DW) technique. We used a mixed methods approach to pilot test the app with 23 first-time orthopaedic visit patients presenting with joint pain and/or function deficiency by presenting the app for utilization and conducting qualitative interviews and quantitative surveys post utilization. Results: The study validated five core TOP domains, with most users dividing their 100-point DW allocation across 1-3 domains. The tool received moderate to high usability scores. Thematic analysis of patient interviews provides insights into TOPs that are important to patients, how they can be communicated effectively, and incorporated into a clinical visit with meaningful patient-provider communication that leads to shared decision making. Conclusion: Patient TOPs may be important human factors to consider in determining treatment options that may be helpful for automating patient treatment recommendations. We conclude that inclusion of patient TOPs to inform the design of HRSs results in creating more robust patient treatment profiles in the EHR thus enhancing opportunities for treatment recommendations and future AI applications.

Uses of Health Care System Medical Care Services by Athletes After Injury at the High School Level.

BACKGROUND: Health care utilization can vary by age group, geographic location, and socioeconomic status (SES). A paucity of information exists regarding the availability and utilization of medical care by injured scholastic athletes. The purpose of this study was to describe and compare injuries and health care service utilization by school SES over an academic year. METHODS: Injury and health care service data was collected from 1 large school district. Percentage of free and reduced lunch (FRPL) for each school was calculated to stratify schools into high (<50% FRPL) and low (≥50.1% FRPL) SES groups. Incidence proportion and relative risk (RR) with 95% confidence intervals (95% CI) were calculated. RESULTS: About 1756 injuries were reported among over 7000 participating athletes from 14 high schools. Similar injury incidence proportions were reported between high and low SES schools (RR = 1.10 [1.00-1.20]). Athletes from low SES schools were twice (RR = 2.01 [1.21-3.35]) and over three (RR = 3.42 [1.84-6.55]) times more likely to receive emergency and physical therapy care. SES was not associated with the use of physician, imaging, or surgery services. IMPLICATIONS FOR SCHOOL HEALTH, POLICY, AND EQUITY: School medical providers and administrators should have ready and provide a list of trusted outside primary care and specialty providers that have experience in sports medicine. They should also enquire and follow up on which outside provider the high school athlete will seek care when referring out to outside providers. CONCLUSIONS: Injury incidence was similar between high and low SES schools. However, athletes from low SES high schools were over 2-fold more likely to use emergency department services. Understanding factors influencing health care services choice and usage by student athletes from different socioeconomic backgrounds may assist sport medicine clinicians in identifying barriers and potential solutions in improving time to health restoration, athlete outcomes, and health care monetary burden.

Comparison of instrumental variable analysis using a new instrument with risk adjustment methods to reduce confounding by indication.

Confounding by indication is a vexing problem, especially in evaluating treatment effects using observational data, since treatment decisions are often related to disease severity, prognosis, and frailty. To compare the ability of the instrumental variable (IV) approach with a new instrument based on the local-area practice style and risk adjustment methods, including conventional multivariate regression and propensity score adjustment, to reduce confounding by indication, the authors investigated the effects of long-term control (LTC) therapy on the occurrence of acute asthma exacerbation events among children and young adults with incident and uncontrolled persistent asthma, using Iowa Medicaid claims files from 1997-1999. Established evidence from clinical trials has demonstrated the protective benefits of LTC therapy for persistent asthma. Among patients identified (n = 4,275), those with higher asthma severity at baseline were more likely to receive LTC therapy. The multivariate regression and propensity score adjustment methods suggested that LTC therapy had no effect on the occurrence of acute exacerbation events. Estimates from the new IV approach showed that LTC therapy significantly decreased the occurrence of acute exacerbation events, which is consistent with established clinical evidence. The authors discuss how to interpret estimates from the risk adjustment and IV methods when the treatment effect is heterogeneous.

Apples and oranges? Interpretations of risk adjustment and instrumental variable estimates of intended treatment effects using observational data.

Instrumental variable (IV) and risk adjustment (RA) estimators, including propensity score adjustments, are both used to alleviate confounding problems in nonexperimental studies on treatment effects, but it is not clear how estimates based on these 2 approaches compare. Methodological considerations have shown that IV and RA estimators yield estimates of distinct types of causal treatment effects regardless of confounding problems. Many investigators have neglected these distinctions. In this paper, the authors use 3 schematic models to explain visually the relations between IV and RA estimates of intended treatment effects as demonstrated in the methodological studies. When treatment effects are homogeneous across a study population or when treatment effects are heterogeneous across the study population but treatment decisions are unrelated to the treatment effects, RA and IV estimates should be equivalent when the respective assumptions are met. In contrast, when treatment effects are heterogeneous and treatment decisions are related to the treatment effects, RA estimates of treatment effect can asymptotically differ from IV estimates, but both are correct even when the respective assumptions are met. Appropriate interpretations of IV or RA estimates can be facilitated by developing conceptual models related to treatment choice and treatment effect heterogeneity prior to analyses.