Stratum-specific Likelihood Ratio Research Articles

Application of a stratum-specific likelihood ratio analysis in a screen for depression among a community-dwelling population in Japan.

BackgroundEfficient screening for depression is important in community mental health. In this study, we applied a stratum-specific likelihood ratio (SSLR) analysis, which is independent of the prevalence of the target disease, to screen for depression among community-dwelling individuals.MethodThe Center for Epidemiologic Studies Depression Scale (CES-D) and the Mini International Neuropsychiatric Interview (MINI) were administered to 789 individuals (19–87 years of age) who participated in the Iwaki Health Promotion Project 2011. Major depressive disorder (MDD) was assessed using the MINI.ResultsFor MDD, the SSLRs were 0.13 (95% CI 0.04–0.40), 3.68 (95% CI 1.37–9.89), and 24.77 (95% CI 14.97–40.98) for CES–D scores of 0–16, 17–20, and above 21, respectively.ConclusionThe validity of the CES-D is confirmed, and SSLR analysis is recommended for its practical value for the detection of individuals with the risk of MDD in the Japanese community.

Evidence-based thresholds for the volume-value relationship in shoulder arthroplasty: outcomes and economies of scale

Whereas several studies suggest that high-volume surgeons and hospitals deliver superior patient outcomes with greater cost efficiency, no evidence-based thresholds separating high-volume surgeons and hospitals from those that are low or medium volume exist in shoulder arthroplasty. The objective of this study was to establish meaningful thresholds that take outcomes and cost into consideration for surgeons and hospitals performing shoulder arthroplasty. Using 9546 patients undergoing primary shoulder arthroplasty for osteoarthritis from an administrative database, we created and applied 4 models using stratum-specific likelihood ratio (SSLR) analysis of a receiver operating characteristic (ROC) curve. We generated 4 sets of thresholds predictive of adverse outcomes, namely, increased length of stay (LOS) and increased cost for both surgeon and hospital volume. SSLR analysis of the 4 ROC curves by surgeon volume produced 3 volume categories. LOS and cost by annual shoulder arthroplasty surgeon volume produced the same strata: 0-4 (low), 5-14 (medium), and 15 or more (high). LOS and cost by annual shoulder arthroplasty hospital volume produced the same strata: 0-3 (low), 4-14 (medium), and 15 or more (high). LOS and cost decreased significantly (P < .05) in progressively higher volume categories. Our study validates economies of scale in shoulder arthroplasty by demonstrating a direct relationship between volume and value through SSLR analysis of ROC curves for risk-based volume stratification using meaningful volume definitions for low-, medium-, and high-volume surgeons and hospitals. The described volume-value relationship offers patients, surgeons, hospitals, and other stakeholders meaningful thresholds for the optimal delivery of shoulder arthroplasty.

The Relation Between Volume of ACL Reconstruction and Future Knee Surgery

Objectives:A large body of volume-outcomes literature has now demonstrated a strong positive relationship between increasing annual surgical volume and improved patient outcomes across a wide variety of surgical procedures, including orthopedics. Unfortunately, most previous research has been limited by methods that arbitrarily assign cutoffs of “low” or “high” volume that may not be clinically meaningful. Thus, it is unknown how many cases should be performed each year to maintain competence in the procedure. The purpose of this study was to determine meaningful annual volume thresholds for ACL reconstruction.Methods:We identified ACL reconstructions performed in New York (NY) State hospitals between 2003 and 2014 using the NY-SPARCS hospital in-patient and ambulatory surgery database. These cases were followed until subsequent ipsilateral knee surgery (revision ACL, meniscus/cartilage surgery, or total knee replacement) or until the end of the study period, December 31, 2014. Surgeon volume was calculated as the number of ACL reconstructions performed by that surgeon in the 365 days prior to the case of interest. Stratum specific likelihood ratio (SSLR) analysis is a method to identify clinically meaningful cutpoints by correlating volume with outcomes through evaluating a receiver operating characteristic (ROC) curve. A Cox proportional hazards model was used to measure the effect of surgeon annual ACL reconstruction volume on risk of subsequent ipsilateral knee surgery adjusting for patient characteristics: age, sex, race, and insurance type.Results:Between 2003 and 2014, 77,899 ACL reconstructions were performed in NY State by 1,316 surgeons. Mean patient age was 30.8+/- 12.5 years, and patients were 61% male, 65% white race, and 74% covered by private insurance. SSLR analysis revealed 2 meaningful cutpoints in risk of subsequent ipsilateral knee surgery: 17 & 35 cases per year. The Cox proportional hazards model demonstrated a 29% decreased risk of ipsilateral knee surgery for those ACL reconstructions by surgeons performing >35 cases per year compared to those performing <17 (Hazard Ratio [HR] 1.29, 95% confidence interval [CI] 1.23-1.35, p<0.001). Further, surgeons in the mid-range group performing 18-35 ACL reconstructions per year demonstrated a 6% decreased risk of subsequent ipsilateral knee surgery (HR 1.06, 95% CI 1.01-1.11, p=0.014).Conclusion:This study identified three ranges for ACL reconstruction volume that result in different clinical outcomes. The highest cutpoint was 35 ACL reconstructions per year, which resulted in the lowest percentage of future knee surgery. These findings suggest orthopedic surgeons should perform a minimum of 35 ACL reconstructions per year to maintain clinical competence with the procedure.

Meaningful Thresholds for the Volume-Outcome Relationship in Total Knee Arthroplasty.

Increasing evidence supports the finding that patients undergoing a total knee arthroplasty with high-volume physicians and hospitals achieve better outcomes. Unfortunately, the existing definitions for high-volume surgeons and hospitals are highly variable and entirely arbitrary. The aim of this study was to identify a set of meaningful hospital and surgeon total knee arthroplasty volume thresholds. Using 289,976 patients undergoing primary total knee arthroplasty from an administrative database, we applied stratum-specific likelihood ratio (SSLR) analysis of a receiver operating characteristic (ROC) curve to generate sets of volume thresholds most predictive of adverse outcomes. The outcomes considered for surgeon volume included 90-day complication and 2-year revision. For hospital volume, we considered 90-day complications and 90-day mortality. SSLR analysis of the ROC curves for 90-day complication and 2-year revision rates by surgeon volume identified four volume categories: 0 to 12, 13 to 59, 60 to 145, and ≥146 total knee arthroplasties per year. Complication rates decreased significantly (p < 0.05) in progressively higher-volume categories. Revision rates followed a similar pattern, but did not decrease between surgeons performing 60 to 145 arthroplasties per year and those performing ≥146 arthroplasties per year. SSLR analysis of 90-day complication and 90-day mortality rates by hospital volume also identified four volume categories: 0 to 89, 90 to 235, 236 to 644, and ≥645 total knee arthroplasties per year. Complication rates decreased significantly (p < 0.05) in progressively higher-volume categories, but the rates did not decrease between hospitals performing 236 to 644 arthroplasties per year and those performing ≥645 arthroplasties per year. Mortality rates for hospitals with ≥645 total knee arthroplasties per year were significantly lower (p < 0.05) than those below the threshold. Our study supports the use of SSLR analysis of ROC curves for risk-based volume stratification in total knee arthroplasty volume-outcomes research. SSLR analysis established meaningful volume definitions for low, medium, high, and very high-volume total knee arthroplasty surgeons and hospitals. This should help patients, surgeons, hospitals, and policymakers to make decisions with regard to the optimal delivery of total knee arthroplasty. Therapeutic Level III. See Instructions for Authors for a complete description of levels of evidence.

Reliability and validity of the new Tanaka B Intelligence Scale scores: a group intelligence test.

ObjectiveThe present study evaluated the reliability and concurrent validity of the new Tanaka B Intelligence Scale, which is an intelligence test that can be administered on groups within a short period of time.MethodsThe new Tanaka B Intelligence Scale and Wechsler Intelligence Scale for Children-Third Edition were administered to 81 subjects (mean age ± SD 15.2±0.7 years) residing in a juvenile detention home; reliability was assessed using Cronbach’s alpha coefficient, and concurrent validity was assessed using the one-way analysis of variance intraclass correlation coefficient. Moreover, receiver operating characteristic analysis for screening for individuals who have a deficit in intellectual function (an FIQ<70) was performed. In addition, stratum-specific likelihood ratios for detection of intellectual disability were calculated.ResultsThe Cronbach’s alpha for the new Tanaka B Intelligence Scale IQ (BIQ) was 0.86, and the intraclass correlation coefficient with FIQ was 0.83. Receiver operating characteristic analysis demonstrated an area under the curve of 0.89 (95% CI: 0.85–0.96). In addition, the stratum-specific likelihood ratio for the BIQ≤65 stratum was 13.8 (95% CI: 3.9–48.9), and the stratum-specific likelihood ratio for the BIQ≥76 stratum was 0.1 (95% CI: 0.03–0.4). Thus, intellectual disability could be ruled out or determined.ConclusionThe present results demonstrated that the new Tanaka B Intelligence Scale score had high reliability and concurrent validity with the Wechsler Intelligence Scale for Children-Third Edition score. Moreover, the post-test probability for the BIQ could be calculated when screening for individuals who have a deficit in intellectual function. The new Tanaka B Intelligence Test is convenient and can be administered within a variety of settings. This enables evaluation of intellectual development even in settings where performing intelligence tests have previously been difficult.

The Alcohol Use Disorders Identification Test-Consumption (AUDIT-C) in the Assessment of Alcohol Use Disorders Among Acute Injury Patients

The Alcohol Use Disorders Identification Test-Consumption (AUDIT-C) is a brief alcohol screening test and a candidate for inclusion in recommended screening and brief intervention protocols for acute injury patients. The objective of the current study was to examine the performance of the AUDIT-C to risk stratify injury patients with regard to their probability of having an alcohol use disorder. Participants (n = 1,004) were from a multisite Australian acute injury study. Stratum-specific likelihood ratio (SSLR) analysis was used to examine the performance of previously recommended AUDIT-C risk zones based on a dichotomous cut-point (0 to 3, 4 to 12) and risk zones derived from SSLR analysis to estimate the probability of a current alcohol use disorder. Almost a quarter (23%) of patients met criteria for a current alcohol use disorder. SSLR analysis identified multiple AUDIT-C risk zones (0 to 3, 4 to 5, 6, 7 to 8, 9 to 12) with a wide range of posttest probabilities of alcohol use disorder, from 5 to 68%. The area under receiver operating characteristic curve (AUROC) score was 0.82 for the derived AUDIT-C zones and 0.70 for the recommended AUDIT-C zones. A comparison between AUROCs revealed that overall the derived zones performed significantly better than the recommended zones in being able to discriminate between patients with and without alcohol use disorder. The findings of SSLR analysis can be used to improve estimates of the probability of alcohol use disorder in acute injury patients based on AUDIT-C scores. In turn, this information can inform clinical interventions and the development of screening and intervention protocols in a range of settings.

Maternity blues as predictor of postpartum depression: A prospective cohort study among Japanese women

Background. Maternity blues and postpartum depression are common complications of childbearing. However, few studies have shown the relationship between the severity of maternity blues and the risk of postnatal depression. We carried out a longitudinal study among Japanese women to show that maternity blues is a useful factor for predicting postpartum depression.Methods. Two hundred and thirty-five women completed questionnaires before delivery, and five days, one month, and three months after delivery. They were required to answer the Stein's Blues Scale and Edinburgh Postnatal Depression Scale and other variables. A sequential logistic regression analysis was performed to estimate the association of maternity blues with postpartum depression. The stratum-specific likelihood ratio was then calculated.Results. The prevalence of postpartum depression was 12.8%. A Stein's Blues Scale of eight or above was significantly associated with postnatal depression. Likelihood ratios (95% CI ) for the Stein's Blues Scale of 0 to 3, 4 to 7, 8 to 11 and 12 or more were 0.33 (0.16–0.65), 1.06 (0.60–1.88), 3.42 (1.64–7.12) and 9.57 (3.41–26.86), respectively.Conclusions. These findings suggest that maternity blues is a strong predictor of postpartum depression. The higher the blues score, the higher the risk of postpartum depression.

선택편향이 존재할 때, 수정 층화우도비를 이용한 최적절사점의 결정

Stratum-specific likelihood ratio, which is ratio of the sensitivity to 1-the specificity in each stratum of the test, could be biased if the sensitivity and specificity of diagnostic test are affected by verification bias. Therefore, the optimal cutoff point determined by biased stratum-specific likelihood ratios is incorrect. In this study, we derived adjusted stratum-specific likelihood ratios using the adjusted sensitivity and specificity, and obtained the adjusted optimal cutoff point. The influence of the verification bias on the optimal cutoff point was described through the relation between adjusted and unadjusted stratum-specific likelihood ratios.

Response to Bansal and Schwartz

TO THE EDITOR: We thank Dr. Bansal and Dr. Schwartz for their interest in our study. We used the stratum-specific likelihood ratio method described by Peirce and Cornell in an exploratory analysis to determine the optimum cutoff for serum ferritin, and not simply to compute a likelihood ratio for ferritin > 100 ng/mL (1). Therefore, only stratum-specific likelihood ratio was defined in the methods section of the article (2). In our experience, accuracy, positive predictive value, and negative predictive value are better comprehended by patients when discussing risks and benefits of a diagnostic test. These are reported in Table 5 of the article (2).

Application of the stratum-specific likelihood ratio (SSLR) analysis to results of a depressive symptoms screening survey among Japanese workers

The stratum-specific likelihood ratio (SSLR) has been widely used as a convenient method to interpret the results of screening surveys. In the present paper, an SSLR analysis is applied to the results of a survey that used the Center for Epidemiologic Studies Depression Scale (CES-D) to identify depressive symptoms in a group of Japanese workers in an attempt to determine score categories that have predictive clinical values in a screening context. The self-reported CES-D and the Mini International Neuropsychiatric Interview (MINI) were administered to 2,219 workers (84.2% men; age 21-68 years) during a periodical medical examination. The SSLR was calculated to determine strata with optimal discrimination for major depressive disorder (MDD). For MDD, the SSLR was 0.06 (95%CI: 0.02-0.18) for the CES-D score range 0-16, 1.90 (0.78-4.62) for the score range 17-19, and 12.4 (10.2-15.1) for the score above 20. The SSLR shown in this study proposes score categories that guide clinicians in selecting the workers most at risks of having MDD among Japanese workers.

Adult-age donors offer acceptable long-term survival to pediatric heart transplant recipients: An analysis of the United Network of Organ Sharing database

A critical shortage of donor organs has caused many centers to use less restrictive donor criteria, including the use of adult-age donors for pediatric recipients. The purpose of this study is (1) to describe the supply of pediatric (0-18 years) heart donors, (2) to explore the relationship between donor age and long-term survival, and (3) to define threshold age ranges associated with decreased long-term survival. The United Network of Organ Sharing provided deidentified patient-level data. Primary analysis focused on 1887 heart transplant recipients aged 9 to 18 years undergoing transplantation from October 1, 1987, to September 25, 2005. Kaplan-Meier analysis and log-rank tests were used in time-to-event analysis. Receiver operating characteristic curves and stratum-specific likelihood ratios were generated to compare survival at various donor age thresholds. The number of pediatric donors decreased (P < .001) over the study period, particularly from 1993 (n = 640) through 2004 (n = 432). Among recipients aged 9 to 18 years, univariate analysis demonstrated a statistically significant (P < .001) inverse relationship between donor age and survival. Stratum-specific likelihood ratio analysis generated 3 strata for donor age: the low-risk, intermediate-risk, and high-risk groups consisted of donors aged 13 years or younger (n = 611, 32.41%), 14 to 51 years (n = 1258, 66.7%), and 52 years and older (n = 16, 0.85%), respectively. In the low-risk, intermediate-risk, and high-risk groups median survival was 4069 days (11.1 years), 3495 days (9.57 years), and 1197 days (3.28 years), respectively. Although donors aged 13 years or less offer pediatric recipients the best chance for achieving long-term survival, donors aged 14 to 51 years offer good outcomes to pediatric recipients. Consideration should be given to expanded use of well-selected adult-age donors for pediatric recipients.

New Approach to the Validity of the Alcohol Use Disorders Identification Test: Stratum-Specific Likelihood Ratios Analysis

Previous validation studies of the Alcohol Use Disorders Identification Test (AUDIT) used either the fixed-threshold approach or receiver operating characteristic (ROC) analyses. This study was aimed to conduct a stratum-specific likelihood ratio (SSLR) analysis on the validity of the AUDIT for detecting alcohol use disorders. A two-phase validation study was conducted among nonpsychiatric inpatients at a general hospital, using the AUDIT for phase 1 screening and the Schedules for Clinical Assessment in Neuropsychiatry for the phase 2 interview. All of the patients with positive screening results and a subsample of the patients with negative screening results were interviewed in the second phase. The validity of the AUDIT for current alcohol use disorders was estimated using the ROC first, and an SSLR analysis was performed subsequently. In 422 inpatients who completed the screening, 193 had a negative result with a zero score, 103 had a negative result with a nonzero score, and 126 had a positive result. Twenty screened patients with a negative result and a nonzero score and 107 patients with a positive result were interviewed in the second phase, in which 59 patients were found to have current alcohol use disorders according to DSM-IV criteria. The area under the ROC curve of the AUDIT for current alcohol use disorders was 0.98. The sensitivity and specificity of the AUDIT at the optimal cutoff of 8 were 0.97 and 0.90, respectively. Three strata of the AUDIT scores (0-7, 8-13, and > or = 14) were derived for alcohol use disorders, with the SSLRs being 0.04, 3.67, and 49.72, and the corresponding posttest probabilities being 0.01, 0.42, and 0.91, respectively. The AUDIT is a valid screening tool for alcohol use disorders, and the three strata derived from the SSLR analysis are informative and readily applicable in clinical practice.

Using the CES-D in a two-phase survey for depressive disorders among nonreferred adolescents in Taipei: a stratum-specific likelihood ratio analysis

Background: Efficient screening is important in two-phase surveys. We examined whether the Center for Epidemiologic Studies Depression Scale (CES-D) is an appropriate instrument for screening depressive disorders among adolescents in Taipei, an ethnic group whose depression is little known of. Methods: Among the 2440 students of a school in Taipei, aged 12 to 16 years who completed the CES-D and eating attitude test (EAT; response rate=98.5%), 178 were randomly selected from four subgroups defined by the 90th percentile of the two screening tools for face-to-face interview, using the Schedule for Affective Disorders and Schizophrenia for Children (K-SADS). Discriminatory validities of instruments for depressive disorders were estimated, and then a stratum-specific likelihood ratio (SSLR) analysis was conducted for instruments with sufficient validity. Results: The prevalence estimates of depressive disorders varied with different levels of impairment, with a value of 2.4% for major depressive disorder and 0.3% for dysthymic disorder if at least two impairment items were endorsed. The areas under the receiver operative characteristic (ROC) curves were consistently high (0.88–0.90) for major depressive disorder with or without impairment requirement, but low (0.49) for dysthymic disorder without impairment requirement. Three strata of CES-D scores (0–28, 29–48, and ≥49) were derived for major depressive disorder with (SSLR=0.63, 3.00, and 11.75) and without (SSLR=0.61, 5.09, and 10.42) impairment requirement. Conclusions: The CES-D is useful in screening for major depressive disorder among nonreferred adolescents. Three strata are recommended for its practical application.