David Carnahan

A Comparison of Obesity Prevalence: Military Health System and United States Populations, 2009–2012

Overweight and obesity prevalence has increased over the past 30 years. Few studies have looked at the enrolled Military Health System (MHS) population (2.2 million per year). This descriptive study examined trends in overweight and obesity in both children and adults from fiscal years 2009 to 2012 and compared them to the U.S. population. Prevalence in MHS children decreased over time for overweight (14.2-13.8%) and obesity (11.7-10.9%). Active duty adults showed an increase in overweight prevalence (52.7-53.4%) and a decrease in obesity prevalence (18.9-18.3%). For nonactive duty, both overweight and obesity prevalence remained relatively unchanged around 33%. For both children and adults, overweight and obesity prevalence increased with age, except for obesity in the nonactive duty ≥ 65 subgroup. When compared to the United States by gender and age, MHS children generally had a lower overweight and obesity prevalence, active duty adults had higher overweight and lower obesity prevalence, and nonactive duty adults had comparable overweight and obesity prevalence, except for obesity in both men in the 40 to 59 subgroup and women in ≥ 60 subgroup. More research on the MHS population is needed to identify risk factors and modifiable health behaviors that could defeat the disease of obesity.

Abdominal circumference is superior to body mass index in estimating musculoskeletal injury risk.

PURPOSE The purpose of this study was to compare body mass index (BMI) and abdominal circumference (AC) in discriminating individual musculoskeletal injury risk within a large population. We also sought to determine whether age or sex modulates the interaction between body habitus and injury risk. METHODS We conducted a retrospective cohort study involving 67,904 US Air Force personnel from 2005 to 2011. Subjects were stratified by age, sex, BMI, adjusted BMI, and AC. New musculoskeletal injuries were recorded relative to body habitus and time elapsed from the start of study. RESULTS Cox proportional hazards regression revealed increased HR for musculoskeletal injury in those with high-risk AC (males, >39 inches; females, >36 inches) compared with HR in those with low-risk AC (males, ≤35 inches; females, ≤32 inches) in all age categories (18-24 yr: HR = 1.567, 95% confidence interval (CI) = 1.327-1.849; 25-34 yr: HR = 2.089, 95% CI = 1.968-2.218; ≥35 yr: HR = 1.785, 95% CI = 1.651-1.929). HR for obese (BMI, ≥30 kg·m) compared with that for normal individuals (BMI, <25 kg·m) were less elevated. Kaplan-Meier curves showed a dose-response relation in all age groups but most prominently in 25- to 34-yr-old participants. Time to injury was consistently lowest in 18- to 24-yr-old participants. Score chi-square values, indicating comparative strength of each model for injury risk estimation in our cohort, were higher for AC than those for BMI or adjusted BMI within all age groups. CONCLUSIONS AC is a better predictor of musculoskeletal injury risk than BMI in a large military population. Although absolute injury risk is greatest in 18- to 24-yr-old participants, the effect of obesity on injury risk is greatest in 25- to 34-yr-old participants. There is a dose-response relation between obesity and musculoskeletal injury risk, an effect seen with both BMI and AC.

Estimating diabetes prevalence in the Military Health System population from 2006 to 2010.

Evidence-based articles have demonstrated an increase in diabetes prevalence, but diabetes prevalence in the enrolled Military Health System population was previously understudied. Variability in diabetes prevalence rates calculated from 5 groups of algorithms was examined in the Military Health System population (3 million enrollees per year) from fiscal years 2006 to 2010. Time trend analysis and rate comparisons to the U.S. population were also performed. Increasing linear trends in diabetes prevalence from 2006 to 2010 were seen in all algorithms, though considerable rate variation was observed within each study year. Prevalence increased with age, except for a slight decrease in those ≥75 years. Overall diagnosed diabetes prevalence ranged from 7.26% to 11.22% in 2006 and from 8.29% to 13.55% in 2010. Prevalence among active duty members remained stable, but a significant upward trend was observed among nonactive duty members across study years. Age-standardized rates among nonactive duty females were higher than the U.S. population rates from 2006 to 2010. This study demonstrates prevalence rate variability because of differing case algorithms and shows evidence of a growing diabetes population in the Military Health System, specifically within the nonactive duty 45 years and older demographic groups. Further research of this population should focus on validation of case definitions.

Metabolic Syndrome in the Military Health System Based on Electronic Health Data, 2009–2012

Metabolic syndrome prevalence in the United States rose from 27% to 34.2% between 1999-2000 and 1999-2006. However, prevalence has not been determined in the Military Health System. This retrospective descriptive study included enrolled Military Health System adults during fiscal years 2009-2012. We explored three populations (nonactive duty, active duty, and Air Force active duty) and their metabolic syndrome components (body mass index or waist circumference, blood glucose test, triglyceride, high density lipoprotein, and blood pressure). The active duty sample (who had all five components measured) was representative of its population, but the nonactive duty sample was not. Therefore, we reported component-wise prevalence for both nonactive and active duty populations, but only reported prevalence of metabolic syndrome for active duty. A decreasing trend, greater in men, was seen. Crude prevalence in 2012 was higher among men and highest among males and females aged 45-64. Only Air Force active duty data contained waist circumference measurements, enabling comparison to the United States. This subgroup prevalence was significantly lower than the United States prevalence in 2010 for both genders in every age group. Although decreasing metabolic syndrome prevalence is promising, prevalence is still high and future research should explore policies to help lower the prevalence.

Opioid Use Patterns Among Active Duty Service Members and Civilians: 2006–2014

Introduction Between 2001 and 2009, opioid analgesic prescriptions in the Military Health System quadrupled to 3.8 million. The sheer quantity of opioid analgesics available sets the stage for issues related to misuse, abuse, and diversion. To address this issue, the Department of Defense implemented several directives and clinical guidelines to improve access to appropriate pain care and safe opioid prescribing. Unfortunately, little has been done to characterize changing patterns of opioid use in active duty service members (ADSM), so little is known about how combat operations and military health care policy may have influenced this significant problem. We examined changes in opioid use for ADSM between 2006 and 2014, compared trends with the civilian population, and explored the potential role of military-specific factors in changes in opioid use in the Military Health System. Materials and Methods After obtaining Institutional Review Board approval, administrative prescription records (Pharmacy Data Transaction Records) for non-deployed ADSM were used to determine the number of opioid prescriptions dispensed each year and the proportion of ADSM who received at least one prescription per month between 2006 and 2014. Based on the observation and the literature, we identified December 2011 as the demarcation point (the optimal point to identify the downturn in opioid use) and used it to compare opioid use trends before and after. We used an autoregressive forecast model to verify changes in opioid use patterns before and after 2011. Several interrupted time series models examined whether military system-level factors were associated with changes in opioid use. Results Between 2006 and 2014, 1,516,979 ADSM filled 7,119,945 opioid prescriptions, either in military treatment facilities or purchased through TRICARE. Both active duty and civilian populations showed signs of decreasing use after 2011, but this change was much more pronounced among ADSM. The forecast model showed a significant difference after 2011 between the projected and actual proportion of ADSM filling an opioid prescription, confirming 2011 as a point of divergence in opioid use. Interrupted time series models showed that the deflection point was associated with significant decreases. A significant increase of 0.261% in opioid prescriptions was seen for every 1,000 wounded in action service members in a given month. Troops returning from Operation Enduring Freedom, Operation Iraqi Freedom, or Operation New Dawn did not appear to influence the rates of use. Even after accounting for returning troops from Operation Enduring Freedom/Operation Iraqi Freedom/Operation New Dawn and wounded in action counts, the deflection point was associated with a lower proportion of ADSM who filled an opioid prescription, leading to a decrease of 1.61% by the end of the observation period (December 2014). Conclusion After December 2011, opioid use patterns significantly decreased in both civilian and ADSM populations, but more so in the military population. Many factors, such as numbers of those wounded in action and the structural organization of the Military Health System, may have caused the decline, although more than likely the decrease was influenced by many factors inside and outside of the military, including policy directives and cultural changes.

Public Health Surveillance via Template Management in Electronic Health Records: Tri-Service Workflow's Rapid Response to an Infectious Disease Crisis

OBJECTIVE To describe the use of template-based screening for risk of infectious disease exposure of patients presenting to primary care medical facilities during the 2014 West African Ebola virus outbreak. METHODS The Military Health System implemented an Ebola risk-screening tool in primary care settings in order to create early notifications and early responses to potentially infected persons. Three time-sensitive, evidence-based screening questions were developed and posted to Tri-Service Workflow (TSWF) AHLTA templates in conjunction with appropriate training. Data were collected in January 2015, to assess the adoption of the TSWF-based Ebola risk-screening tool. RESULTS Among encounters documented using TSWF templates, 41% of all encounters showed use of the TSWF-based Ebola risk-screening questions by the fourth day. The screening rate increased over the next 3 weeks, and reached a plateau at approximately 50%. CONCLUSIONS This report demonstrates the MHS capability to deploy a standardized, globally applicable decision support aid that could be seen the same day by all primary care clinics across the military health direct care system, potentially improving rapid compliance with screening directives.

Water, Water, Everywhere, and Not a Drop to Drink

In themovie Life of Pi, the protagonist and his tiger companion facea slowdeath fromthirstwhile surroundedbyanocean of water. The ubiquity of the desired and vital substance, though inunacceptable form, supplies rich irony to theprotagonist’s struggle, andso it is with the informationneedsof clinicians. It is no great revelation to say we live in a society that carries auniverse of possibilities in its pocket: global telecommunicationandvideoconferencing, instantaneousknowledge at our fingertips, business transactions viamobile applications, and libraries of books,music, andvideos ondemand. Even so, the informationneeds of clinicianswhoprovide care in the office or at the bedside are not so easily met. The systematic reviewbyDelFiol et al1 highlights the continued need for information by clinicians despite the explosion ofmedical literature. It has been estimated that approximately 10 million articles were published between 1865 and 2006, with a steady annual increase in numbers.2 Despite the staggering availability of information,Del Fiol et al found that on average a clinicianwill have 1 question for every 2 patients seen but will seek answers to only 50% of questions and find answers for 80%of questions pursued.How is it that approximately 50% of clinical questions are not acted on?What does it tell us thatwhenquestions are pursued, the clinician is able to find an answer 80% of the time? The answers to these questions probably lie in the barriers summarized byDel Fiol et al.1 The lack of time at the point of care and subsequent forgetting of questions could explain whymany clinicians do not attempt to answer questions that arise from patient care, and the demands of caring for a high volume of patients and the attendant after-hours work could explainwhymany clinicians defer searching for answers. It is possible, however, that questions disregarded as insignificant could changepatient care if information that includedanswerswasprovided seamlessly. Fortunately, several developments in recent years helppromotemore seamless delivery of information at the point of care. In 2011, phase 1 of Meaningful Use, a product of the Health Information Technology for Economic and Clinical Health Act, incentivized and mandated the adoption of certified electronic health records (EHRs). This certification requires EHRs to use established medical terminology coding standards for medications, laboratories, and diagnoses (eg, RxNorm, LOINC [Logical Observation Identifiers Names and Codes], and SNOMED CT [Systematized Nomenclature of Medicine–Clinical Terms]), which is key to fostering interoperability between clinical information systems and, eventually, health information sharing. As noted by Del Fiol et al,1 the standard used to send messages from one system to another (Health Level Seven) has included a contextaware knowledge retrieval “Infobutton” as part of its standard, which has also been included among the requirements of EHR certification.3 Unfortunately, this process is only part of the equation. Used alone, it would be as if the protagonist Pi developed an apparatus that could extract water from the ocean but not convert salt water into fresh water; the result is still no potable water. The knowledge resources (eg, PubMed,Medline Plus,National ClearinghouseGuidelines, andUpToDate) are the other half of the equation. Once a clinical topic is identified via one of the common standards found in the EHR, the topic then needs tobemapped to the semantic taxonomyusedbyknowledge resource developers. In most cases, knowledge resourcedevelopersuseMeSH(MedicalSubjectHeadings) terms, Unified Medical Language System, or nonstandardized textbased classifications,which arenot necessarily goodmatches to the standard codes from theEHR.4 The ability tomatch the information needed depends on the quality and detail of the metadata surrounding the information resources aswell as the level of specificity requiredby thequestion. This presents the challenge of knowing howmuch information to provide and howtoprovide it. Thehuman-computer interactionwithin information-seeking behaviors needs to be better understood if information is to be provided unobtrusively, and this consideration could affect future adoption. Fortunately, the Clinical Decision Support Consortium is working to promote an Enterprise Clinical Decision Support Framework, which would ideally support information services in the formof clinical guideline alerts in theEHR.5Moreover, the Clinical Decision Support Work Group’s effort with Infobuttons seems uniquely poised for the provision of “context-aware” general information.3 However, these efforts do not seem to encompass the unique patient-specific information required formore complexquestions. Interestingwork is being done on the use of natural language processing to rank key sentences from primary literature abstracts that apply to a specific question asked.6,7 Although this use falls short of an automatedcontext-specificquestion-answer system, it is a significant start toward that end. Imagine the clinical impact if your patient’s EHR provided summary information tohelp youdetermine the risk of startingwarfarin therapy in an octogenarianwith a history of strokeandnew-onset atrial fibrillation, alongwith links to specific articles in the primary literature. Again, this level of detail will require use ofmore robustmetadata or semantic taxonomy on the knowledge resource side to match patient demographics and clinical details, but this approach couldultimately improve the application of study results by matching population characteristics to individual patients, thereby ensuring external validity. The realmessage of the reviewbyDel Fiol et al1 is not that clinicians have questions but that many of our questions remainunanswereddespite theoceanof informationaroundus. Related article page 710 Questions Raised by Clinicians at Point of Care Original Investigation Research