Development of an algorithm to identify fall-related injuries and costs in Medicare data
© Kim et al. 2016
Received: 16 October 2015
Accepted: 29 December 2015
Published: 5 January 2016
Identifying fall-related injuries and costs using healthcare claims data is cost-effective and easier to implement than using medical records or patient self-report to track falls. We developed a comprehensive four-step algorithm for identifying episodes of care for fall-related injuries and associated costs, using fee-for-service Medicare and Medicare Advantage health plan claims data for 2,011 patients from 5 medical groups between 2005 and 2009.
First, as a preparatory step, we identified care received in acute inpatient and skilled nursing facility settings, in addition to emergency department visits. Second, based on diagnosis and procedure codes, we identified all fall-related claim records. Third, with these records, we identified six types of encounters for fall-related injuries, with different levels of injury and care. In the final step, we used these encounters to identify episodes of care for fall-related injuries.
To illustrate the algorithm, we present a representative example of a fall episode and examine descriptive statistics of injuries and costs for such episodes. Altogether, we found that the results support the use of our algorithm for identifying episodes of care for fall-related injuries. When we decomposed an episode, we found that the details present a realistic and coherent story of fall-related injuries and healthcare services. Variation of episode characteristics across medical groups supported the use of a complex algorithm approach, and descriptive statistics on the proportion, duration, and cost of episodes by healthcare services and injuries verified that our results are consistent with other studies.
This algorithm can be used to identify and analyze various types of fall-related outcomes including episodes of care, injuries, and associated costs. Furthermore, the algorithm can be applied and adopted in other fall-related studies with relative ease.
KeywordsFall Fall-related injury Algorithm Medicare ICD-9-CM HCPCS
Falls and fall-related injuries are costly to society, and most commonly occur in the older population (Tinetti et al. 2008). Although the impact of falls on costs and interventions to reduce fall-related injuries are major areas of ongoing research, identifying fall-related injuries poses a major challenge for researchers since tracking actual falls requires examining medical records or patient self-report augmented with memory aids, such as calendars or diaries (Ray et al. 1992; Rizzo et al. 1996; Rizzo et al. 1998). Medical record review is expensive and self-report may be inaccurate because of recall problems or the hassle of completing memory aids (Ganz et al. 2005). The use of healthcare claims data provides an alternative way of identifying fall-related injuries that is cost-efficient and easier to implement (Bohl et al. 2010; Carter and Porell 2011; Finkelstein et al. 2005; Tinetti et al. 2008). Furthermore, the accuracy of identifying injuries from claims may increase when using complex algorithms instead of simple definitions (Ray et al. 1992).
In addition to their potential utility in capturing fall-related injuries, healthcare claims may be helpful more generally in testing different definitions of episodes of care for purposes of refining alternative approaches to reimbursement, such as bundled payment. The concept of an episode of care has long been the subject of public interest and debate for reforming the healthcare system. From the Medicare Participating Heart Bypass Center Demonstration in 1991 to the recent Bundled Payments for Care Initiative in 2013, efforts have been made to explore episode-based bundled payment systems that would enhance coordination among providers, improve the quality of care, and reduce healthcare expenditures (Chen et al. 2015; Mechanic 2015). Moreover, the Centers for Medicare and Medicaid Services (CMS) has proposed the Comprehensive Care for Joint Replacement program to begin in January 1, 2016, which is a mandatory Medicare bundled payment scheme for hip and knee replacements (Chen et al. 2015; Mechanic 2015). In this study, we build on previous work to develop a novel approach to identifying episodes of care for fall-related injuries.
Empirical studies have used various measures and types of information to define and classify falls or fall-related injuries. Some studies have classified falls or fall-related injuries based on the setting in which injuries are treated, such as acute care hospital or emergency department (ED) visits (Bohl et al. 2010; Finkelstein et al. 2005; Stevens et al. 2006). Others have used more sophisticated definitions that also involve body sites or types of injury such as hip and other fractures, head injuries, and joint dislocations (Ray et al. 1992; Rizzo et al. 1996; Rizzo et al. 1998; Tinetti et al. 2008). Recent studies have incorporated both severity and body location to identify falls and injuries (Bohl et al. 2010; Bohl et al. 2012; Carter and Porell 2011). For example, Carter and Porell (2011) use location as a measure of injury severity to identify hospitalized and non-hospitalized fallers while Bohl et al. (2012) assign an Injury Severity Score (ISS) to the identified injury episodes and use this information to distinguish sentinel injuries from minor injuries.
Real-life evaluations of falls prevention include a complex mix of patients moving between payers, even within Medicare. However, most studies using administrative data have examined claims data from either traditional fee-for-service Medicare, or health maintenance organizations, but not both due to the complexity involved in obtaining and combining the datasets. Studies have typically used general International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis codes as well as fall-related external cause of injury codes (E codes) to identify fall-related claims (Bohl et al. 2010; Ray et al. 1992; Tinetti et al. 2008). E codes can be used to identify fall-related claims where their use is mandated, a state-by-state decision. However, in studies that span several states, reliance on E codes may result in incomplete identification of fall-related injuries and potential misclassification of injured and non-injured individuals.
The outcome of interest also affects how falls or fall-related injuries are defined. For example, specific fractures for each body site can be counted so that a single injury event could involve multiple fractures (Ray et al. 1992). Counts of injuries are often used to calculate occurrence rates (e.g., per 1000 person-years) (Tinetti et al. 2008). Another approach is injury episodes, which include all injuries within a designated window of days (e.g., 90 or 180 days) (Carter and Porell 2011). While some earlier studies used charge data, more recent studies used medical expenditures as a close estimate of fall-related costs (Alexander et al. 1992; Bohl et al. 2010; Carter and Porell 2011; Finkelstein et al. 2005; Rizzo et al. 1998).
However, studies have rarely taken all of these factors into account when using claims data to identify fall-related injuries. And while each study estimates the impact of falls and fall-related injuries, it is difficult to compare the results across studies since they involve different definition of falls, different information, and different assumptions. Therefore, we developed a comprehensive algorithm for identifying episodes of care for fall-related injuries, which we defined as a collection of fall-related injury claims that are clustered closely enough within a specified time period to believably stem from the same underlying injury. In this algorithm, we identified fall-related injuries by their healthcare services setting, body sites, and by types of injury. We also used both fee-for-service (FFS) Medicare data from CMS and Medicare Advantage (MA) health plan data, which required integrating datasets from different sources with variation in data format and availability. (For simplicity, we use the term “claims” to refer to data from both FFS Medicare and MA health plans, even though not all MA data are generated from requests for payment.) In particular, the algorithm involved the use of E codes (when available) as well as general ICD-9-CM diagnosis codes, and identified fall-related injuries as well as episodes of care so that both frequencies and costs of fall-related injuries could be examined in detail.
Our algorithm builds on previous work in several ways. We used work by Taylor et al. and Ray et al. that applied complex algorithms to Medicare data (specifically, inpatient, ED, outpatient, and physician files) to examine incidence of fractures (Ray et al. 1992; Taylor et al. 2011). We adopted the classification of the five types of injuries (hip fractures, other fractures, head injuries, joint dislocations, and fall-related use of medical services (E codes)) from Tinetti et al (2008). We follow Bohl et al. (2012) and also use location as a proxy for severity of the injuries. Our algorithm provides two particularly novel contributions to the field. First, the algorithm is flexible with regard to the sources of claims data; specifically, the algorithm can make use of both FFS Medicare and MA health plan data together. Second, the definition of an episode is more refined, since it is based on the actual pattern of claims rather than a fixed time window, thereby allowing actual care rendered to drive episode duration and thereby capture the appropriate costs.
We developed an algorithm to analyze fall-related outcomes for the ACOVEprime study, which involved a multicomponent intervention in primary care practices to improve the quality of care for falls (Ganz et al. 2015; Wenger et al. 2010). We limited our data analysis to a group of patients age 75 and older with high risk of falling, who were receiving care from practices located in five distinct locations, and covered by either FFS Medicare or MA. We collected Medicare enrollment information and claims data for 2,011 patients from 5 medical groups between 2005 and 2009. We obtained Medicare FFS data, as well as comparable MA datasets from 5 health plans in 2 medical groups (We provide the descriptions of the datasets, data cleaning and combining process, and key variables used in the algorithm in Additional file 1: Appendix A since the information is important for understanding the results as well as for applying the algorithm in a different context).
Using merged Medicare FFS and MA datasets, we developed an algorithm which involved four steps to identify episodes of care for fall-related injuries and associated costs. We describe the algorithm in stepwise fashion to illustrate how we coded and executed the algorithm. First, as a preparatory step, we identified events with care received in acute inpatient and skilled nursing facility (SNF) settings, in addition to ED visits. Second, we identified all records with diagnosis and procedure codes relevant to falls or fall-related fractures. Third, with these records, we defined and identified six different types of encounters for fall-related injuries, based on the healthcare setting and type of codes involved. Fourth, based on these encounters, we identified episodes of care for fall-related injuries.
The first step involved preparatory work for later steps of the algorithm. First, we identified claims and line items for care provided in the inpatient and SNF setting; this information was used in the third step of the algorithm. We identified inpatient and SNF claims for FFS datasets using the short stay/long stay/SNF indicator code variable included in the MedPAR dataset. For MA datasets, which varied in the type and availability of information, we used a case-by-case approach utilizing all available information (see Additional file 1: Appendix B). For both FFS and MA datasets, we also applied a list of Current Procedural Terminology (CPT-4) codes (also known as HCPCS level I codes) that identified inpatient and SNF settings; for identifying the inpatient setting, we used codes 99217–99223, 99231–99236, 99238–99239, 99251–99255, 99261–99263, 99291–99292, and 99356–99357 whereas for the SNF setting we used codes 99301–99313, 99315–99316, and 99318.
Another process in step 1 involved identifying ED visits, independent of inpatient visits, which was used in the fourth step of the algorithm. A claim or line item was considered as an ED visit or related activity if it satisfied one of the following three criteria: (1) Berenson-Eggers Type of Service (BETOS) code, which is based on the HCPCS code, is “M3” or starts with the alphabet letter “I”, (2) place of service code refers to “Emergency Room – Hospital”, or (3) revenue code is 450–459 or 981 (Centers for Medicare & Medicaid Services; Merriman and Caldwell 2003).
The second step of the algorithm involved identifying all fall-related incidents using claims and line items with diagnosis and procedure codes relevant to fractures or fall-related injuries. For diagnosis codes we used all available primary and secondary ICD-9-CM codes, including E codes. For procedure codes we used CPT-4 codes which we further distinguished as repair, casting, splinting, or imaging codes. All fall-related incidents were classified as one of the following five types of injuries: hip fracture, other fracture, head injury, joint dislocation, or E codes for accidental falls. Additional file 1: Appendix C provides a detailed description of the different types of injury and associated diagnosis and procedure codes used.
If a given day did not involve an inpatient or SNF stay and did not qualify as a “(3) probable outpatient fall-related injury” or an “(4) outpatient fracture repair procedure”, we applied the following process. If a given day involved a fall-related diagnosis or an imaging procedure, we looked forward up to 10 days after the given day and searched for another day with a diagnosis code on the same body site or type of injury (as listed in Table C.1 in Additional file 1: Appendix C). If so, we “connected” (i.e., linked) the days of the two sets of records. Using the latter day as the reference date, we repeated this search and connection process until we no longer observed a subsequent day with a relevant diagnosis code within a 10 day window. If at least one connection occurred for a given day, we considered all connected days as a single “(5) possible outpatient fall-related injury.” For patient days that were not identified as any of the encounter types above, we examined whether they involved an E code for accidental falls. If so, we classified these days as “(6) fall-related use of medical care.”
We repeated this process until all fall-related incident days were examined and fall-related injury encounters were identified, where present, for all patients. Based on clinical judgment, we applied the process at differing levels for each of the five types of fall-related injury (that were identified in the second step of the algorithm). Specifically, for “hip fracture” and “other fracture”, we identified encounters (1) to (4). For “other fracture” with associated list of imaging codes and “joint dislocation”, we identified encounters (1) to (5). For “head injury,” we only examined these codes in the inpatient and SNF setting and identified encounters (1) and (2). For “fall-related E codes”, we applied all steps of the flow diagram and identified encounters (1) to (6).
Then, we focused on identifying the end of this episode. Since an episode of care may involve injuries at several body sites at different points of time with short intervals, we first examined whether there was any other subsequent encounter within 30 days of the given encounter (that represented the beginning of a new episode). If there was a subsequent encounter within 30 days, and this encounter did not involve a disqualifying E code, then we connected these encounters together. We repeated this process until we were unable to find any subsequent encounter within 30 days of the latest connected encounter. We also stopped the process if we came across a subsequent encounter with a disqualifying E code. Once this process came to an end, we defined the latest connected encounter as the end of the episode, and defined the connected set of encounters as a new episode of care. We repeated the whole process until all encounters were examined and episodes of care for fall-related injuries were identified, where present, for all patients.
We also sorted each episode into six mutually exclusive and hierarchical categories, which was based on the most severe and reliable level of injury and care involved in each episode. For example, an episode with “(1) inpatient fall-related injury” and “(4) outpatient fracture repair procedure” encounters was considered as an episode of care involving “(1) inpatient fall-related injury.”
Subsequently, we estimated the cost of healthcare services for each episode, which was adjusted to 2009 U.S. dollars using the medical care component of the consumer price index for all urban consumers (Bureau of Labor Statistics). We calculated two types of cost, namely total and attributable cost. Total cost involved all costs during the duration of each episode of care, while attributable cost only included cost of claims for healthcare that had a fall-related diagnosis code.
Results and discussion
First, we present a representative example of an episode of care for fall-related injuries to illustrate way in which an episode of care is identified using the algorithm. Then we examine descriptive statistics of episodes including frequency, duration, and cost, which corroborate that the algorithm is functioning as intended.
Representative example of an episode of care
The bottom four rows of Fig. 3 show the body sites and types of injuries. We see that this episode involved injuries at more than one body site as it involved a primary diagnosis of hip and other (specifically femur, tibia, or fibula) fractures at the beginning of the episode. Although we cannot determine the exact mechanism of injury for the accidental fall, being listed as “other and unspecified,” the presence of fall-related E codes (E888.9) supports our algorithm’s classification of this episode as being fall-related.
Inpatient and SNF stays accounted for 93 % ($ 38,854/$ 41,963) of all costs during this episode, and almost two thirds of all costs were accrued in the first twelve days of the episode, consistent with the clinical expectation that acuity of care is highest in the early phase of an episode. Examination of the remaining claims data for this patient revealed an additional episode of care, 4 months prior to this episode, which involved a same level tripping code for accidental fall (E885.9) and a single-day outpatient visit with radiologic examination of the shoulder. We did not find any fall-related claims in between or after these episodes.
Descriptive statistics for episodes of care
Episodes of care by types of healthcare services setting and injuries involved
Standardized episodes (per 1000 person-years) a
Proportion of episodes, by healthcare services b
Inpatient or SNF stay (Inp. / SNF)
ED visit or ED-related visit (ED)
Non-ED outpatient visit only (Out.)
Proportion of episodes, by injury b
The next four rows display the proportion of episodes involving different types of healthcare services. All episodes involving an inpatient or SNF stay are counted in the fifth row. Episodes that do not involve inpatient or SNF stay but involve an ED visit fall into the sixth row. The rest of the episodes are categorized as episodes with “non-ED outpatient visit only.” A significant proportion of episodes of care did not involve an inpatient or SNF stay (mean of 57 %, ranging from 48 % to 64 % by medical group).
The last four rows are also mutually exclusive categories that show the proportion of episodes involving different types of injuries. Any episodes that involve hip fractures are categorized as “hip fracture.” Episodes that do not involve hip fractures are further categorized by whether they involve any other fractures or not, and are presented in the last two rows, respectively. Here, we see that a significant proportion of episodes of care were non-fracture injuries, with a mean of 47 %, and ranging from 32 % to 51 % by medical group.
Episodes of care by body sites, injury, and types of healthcare services
Ray et al.
Body sites and causes of injury a
Inp. / SNF
ED & Out. b
Other Fracture : Pelvis
Other Fracture : Rib
Other Fracture : Clavicle
Other Fracture : Humerus
Other Fracture : Radius & ulna
Other Fracture : Navicular (scaphoid)
Other Fracture : Hand
Other Fracture : Femur, tibia, fibula
Other Fracture : Patella
Other Fracture : Ankle
Head Injury : Head fracture
Head Injury : Head trauma
Joint Dislocation : Shoulder
Joint Dislocation : Elbow
Joint Dislocation : Wrist
Joint Dislocation : Knee
E Codes : Stairs or steps
E Codes : Ladders or scaffolding
E Codes : Building or structure
E Codes : Hole
E Codes : One level to another
E Codes : Same level - tripping
E Codes : Same level - pushed
E Codes : Other and unspecified
Average duration and cost per episode of care, by types of healthcare services and injuries
Duration (days) b
Cost (2009) b
Episodes by healthcare services a
Inpatient or SNF stay
ED visit or ED-related visit
Non-ED outpatient visit only
Episodes by injury a
Altogether, we found that the results confirm that the algorithm for identifying episodes of care for fall-related injuries is functioning as intended. When we decomposed an episode, we found that the details present a realistic and coherent story of fall-related injuries and healthcare services. Variation of episode characteristics across medical groups supported the use of a complex algorithm approach, and descriptive statistics on the proportion, duration, and cost of episodes by healthcare services and injuries verified that our results are consistent with other studies. Moreover, robustness checks, presented in Additional file 1: Appendix D, showed that our results are not dramatically sensitive to the assumptions used for the algorithm.
Given the financial and time limitations involved in using traditional methods (i.e., patient self-report, medical record review) to identify fall-related injuries and associated costs, practical methods that are cost-efficient and easier to implement are becoming more important. Consequently, we developed a practical approach that employs a comprehensive algorithm to identify episodes of care for fall-related injuries and related costs. Overall, this approach attempts to create a more flexible, generic approach to identifying fall-related injuries and costs that can be applied to existing data describing mixed samples of Medicare patients. This algorithm can be used to analyze claims datasets from both FFS Medicare and MA health plans, which often display wide variation in data format and availability. (Fall-related studies can also save time and money by selectively using Medicare datasets, as shown in Additional file 1: Appendix E.) Also, this algorithm uses four steps to identify fall-related injuries in different healthcare settings, and can be used for extracting detailed information on fall-related injuries and examining associated costs. In particular, using this algorithm for estimating fall-related healthcare services expenditure may lead to reduced measurement bias and improved understanding of fall-related costs. Specifically, it allows for further and detailed analysis of the episodes of care for falls and fall-related injuries. For example, we can examine episodes of care and compare the costs and practices across regions and healthcare providers. We can also dissect the episode and examine the frequency and costs of specific types of injuries (e.g., contusions and lacerations). And while studies are often limited in data and only look into acute care, studies that use this algorithm will also be able to examine the long-term effect of falls and fall-related injuries and their association with other health conditions (Roudsari et al. 2005; Sattin et al. 1990). Future studies can also incorporate ISS if measuring severity is of interest, especially since it is a common practice to use the ISS for measuring severity of injuries in trauma studies (Cryer 2006).
Our approach has several limitations. First, while the approach used here was motivated by the need to identify fall-related injuries where actual tracking of falls was infeasible, one limitation of using the Medicare claims data is that it is dependent upon accurate and consistent coding of information by healthcare providers (Jacobsen et al. 1992; Roudsari et al. 2005; Taylor et al. 2011). Thus, inaccurate information may lead to underestimation or overestimation of episodes, episode duration, and cost of falls and fall-related injuries. Second, the use of E codes in the algorithm could have driven some of the observed variation in outcomes such as non-fracture injury counts, as discussed earlier in Table 1, since E-coding varies by hospitals and EDs as well as by states. Thus, an important area for future research is the impact of local coding practice on fall-related injury ascertainment. Third, due to the limitations of our data source, we did not consider whether a specific fall or fall-related injury was the inciting cause of a new episode or the consequence of another acute illness. Future work should address this concern by simultaneous review of claims and electronic health record data. Distinguishing the various types of fall episodes would strengthen the ability to focus on falls that are likely to respond to evidence-based interventions as opposed to those where the fall is a manifestation of another acute illness (e.g., infection, stroke). Fourth, we were unable to compare the algorithm’s results with actual medical records of falls. Fifth, we did not use ICD-9-CM V codes to identify ongoing fall-related costs when calculating attributable cost, which might have additionally captured small amounts of cost that could have been missed. And sixth, CMS switched to using ICD-10 codes as of October 2015. However, this algorithm will still be a useful resource for future researchers investigating fall-related injuries for the following reasons. First, it will still be relevant and useful for analyzing data collected prior to the switch from ICD-9-CM to ICD-10. Moreover, this algorithm can be generally applied and widely used with modification and update of the diagnosis codes.
Consequently, recommendations for future research and algorithm improvements are as follows: (1) updating diagnosis codes from ICD-9-CM to ICD-10 codes, (2) applying V codes to the algorithm to improve accuracy of measuring attributable cost, (3) validating the algorithm using data with medical records or patient self-reports to identify actual falls, (4) examining the impact of local coding practice on fall-related outcomes, and (5) examining duration and cost of episodes of care for fall-related injuries by type of injury, time period, and type of healthcare services.
- BETOS code:
Berenson-Eggers Type of Service code
Centers for medicare and medicaid services
- CPT-4 code:
Current Procedural Terminology code
- E code:
external cause of injury code
- HCPCS code:
Healthcare Common Procedure Coding System code
International Classification of Diseases, Ninth Revision, Clinical Modification
Injury severity score
Medicare provider analysis and review
Research data assistance center
skilled nursing facility
This project was supported by a grant from the National Institute on Aging (R01AG036776). One of the authors was supported by the UCLA Claude Pepper Older Americans Independence Center funded by the National Institute on Aging (5P30AG028748)33 and the NIH/National Center for Advancing Translational Science (NCATS) UCLA CTSI Grant Number UL1TR000124.
We thank Beate Danielsen, Health Information Solutions, for assistance with data acquisition. We thank Robin Beckman for programming assistance.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
- Alexander BH, Rivara FP, Wolf ME. The cost and frequency of hospitalization for fall-related injuries in older adults. Am J Public Health. 1992;82(7):1020–3.View ArticleGoogle Scholar
- Bohl AA, Fishman PA, Ciol MA, Williams B, Logerfo J, Phelan EA. A longitudinal analysis of total 3-year healthcare costs for older adults who experience a fall requiring medical care. J Am Geriatr Soc. 2010;58(5):853–60. doi:10.1111/j.1532-5415.2010.02816.x.View ArticleGoogle Scholar
- Bohl AA, Phelan EA, Fishman PA, Harris JR. How are the costs of care for medical falls distributed? The costs of medical falls by component of cost, timing, and injury severity. Gerontologist. 2012;52(5):664–75. doi:10.1093/geront/gnr151.View ArticleGoogle Scholar
- Bureau of Labor Statistics. Databases, Tables & Calculators by Subject - Consumer Price Index - All Urban Consumers. http://data.bls.gov/timeseries/CUUR0000SAM?output_view=pct_12mths. Accessed June 2014.
- Carter MW, Porell FW. The effect of sentinel injury on Medicare expenditures over time. J Am Geriatr Soc. 2011;59(3):406–16. doi:10.1111/j.1532-5415.2010.03283.x.View ArticleGoogle Scholar
- Centers for Medicare & Medicaid Services. Berenson-Eggers Type of Service (BETOS). http://www.cms.gov/Medicare/Coding/HCPCSReleaseCodeSets/BETOS.html. Accessed April 2013.
- Chen LM, Meara E, Birkmeyer JD. Medicare’s Bundled Payments for Care Improvement initiative: expanding enrollment suggests potential for large impact. Am J Manag Care. 2015;21(11):814–20.Google Scholar
- Cryer C. Severity of injury measures and descriptive epidemiology. Inj Prev. 2006;12(2):67–8. doi:10.1136/ip.2006.011668.View ArticleGoogle Scholar
- Finkelstein EA, Chen H, Miller TR, Corso PS, Stevens JA. A comparison of the case-control and case-crossover designs for estimating medical costs of nonfatal fall-related injuries among older Americans. Med Care. 2005;43(11):1087–91.View ArticleGoogle Scholar
- Ganz DA, Higashi T, Rubenstein LZ. Monitoring falls in cohort studies of community-dwelling older people: effect of the recall interval. J Am Geriatr Soc. 2005;53(12):2190–4. doi:10.1111/j.1532-5415.2005.00509.x.View ArticleGoogle Scholar
- Ganz DA, Kim S, Zingmond DS, Ramirez KD, Roth CP, Jennings LA, et al. The effect of falls quality improvement program on fall-related injuries. J Am Geriatr Soc. 2015;63(1):63–70.View ArticleGoogle Scholar
- Inaba K, Goecke M, Sharkey P, Brenneman F. Long-term outcomes after injury in the elderly. J Trauma. 2003;54(3):486–91. doi:10.1097/01.TA.0000051588.05542.D6.View ArticleGoogle Scholar
- Jacobsen SJ, Cooper C, Gottlieb MS, Goldberg J, Yahnke DP, Melton 3rd LJ. Hospitalization with vertebral fracture among the aged: a national population-based study, 1986–1989. Epidemiology. 1992;3(6):515–8.View ArticleGoogle Scholar
- Mechanic RE. Mandatory medicare bundled payment--is it ready for prime time? N Engl J Med. 2015;373(14):1291–3. doi:10.1056/NEJMp1509155.View ArticleGoogle Scholar
- Merriman K, Caldwell D. How to Identify Emergency Room Services in the Medicare Claims Data.: Research Data Assistance Center. Minneapolis: University of Minnesota; 2003.Google Scholar
- Ray WA, Griffin MR, Fought RL, Adams ML. Identification of fractures from computerized Medicare files. J Clin Epidemiol. 1992;45(7):703–14.View ArticleGoogle Scholar
- Rizzo JA, Baker DI, McAvay G, Tinetti ME. The cost-effectiveness of a multifactorial targeted prevention program for falls among community elderly persons. Med Care. 1996;34(9):954–69.View ArticleGoogle Scholar
- Rizzo JA, Friedkin R, Williams CS, Nabors J, Acampora D, Tinetti ME. Health care utilization and costs in a Medicare population by fall status. Med Care. 1998;36(8):1174–88.View ArticleGoogle Scholar
- Roudsari BS, Ebel BE, Corso PS, Molinari NA, Koepsell TD. The acute medical care costs of fall-related injuries among the U.S. older adults. Injury. 2005;36(11):1316–22. doi:10.1016/j.injury.2005.05.024.View ArticleGoogle Scholar
- Sattin RW, Lambert Huber DA, DeVito CA, Rodriguez JG, Ros A, Bacchelli S, et al. The incidence of fall injury events among the elderly in a defined population. Am J Epidemiol. 1990;131(6):1028–37.Google Scholar
- Stevens JA, Corso PS, Finkelstein EA, Miller TR. The costs of fatal and non-fatal falls among older adults. Inj Prev. 2006;12(5):290–5. doi:10.1136/ip.2005.011015.View ArticleGoogle Scholar
- Taylor AJ, Gary LC, Arora T, Becker DJ, Curtis JR, Kilgore ML, et al. Clinical and demographic factors associated with fractures among older Americans. Osteoporos Int. 2011;22(4):1263–74. doi:10.1007/s00198-010-1300-8.View ArticleGoogle Scholar
- Tinetti ME, Baker DI, King M, Gottschalk M, Murphy TE, Acampora D, et al. Effect of dissemination of evidence in reducing injuries from falls. N Engl J Med. 2008;359(3):252–61. doi:10.1056/NEJMoa0801748.View ArticleGoogle Scholar
- Wenger NS, Roth CP, Hall WJ, Ganz DA, Snow V, Byrkit J, et al. Practice redesign to improve care for falls and urinary incontinence: primary care intervention for older patients. Arch Intern Med. 2010;170(19):1765–72. doi:10.1001/archinternmed.2010.387.View ArticleGoogle Scholar