Diagnostic Coding of Elder Mistreatment: Results from a National Database of Medicare Advantage and Private Insurance Patients, 2011–17

Abstract

Health care providers may play an important role in detection of elder mistreatment, which is common but under-recognized. We used the Health Care Cost Institute insurance claims database to describe elder mistreatment diagnosis among Medicare Advantage (MA) and private insurance patients in the United States from 2011–2017. We used International Classification of Diseases (ICD) coding to identify cases, examining the impact of transition from ICD-9 (Ninth Revision) to ICD-10 (Tenth Revision), which occurred in October 2015 and added 14 new codes for “suspected” mistreatment. 8,127 patients (0.051% of all aged ≥65), including 6,304 with MA (0.058%) and 1,823 with private insurance (0.026%) received elder mistreatment diagnosis. Transition from ICD-9 to ICD-10 was associated with a small increase in diagnosis rate, with “suspected” codes used in 45.3% of ICD-10 vs. 9.7% of ICD-9 cases. Overall rates remained low. Rates, settings, and types of diagnosis differed between MA and private insurance patients.

Introduction

Health care providers may play an important role in detection of elder mistreatment, a common but under-recognized phenomenon with serious medical and social consequences. Elder mistreatment encompasses behaviors or negligence against an older adult that result in harm or the risk of harm committed by someone in a relationship with an expectation of trust. It also includes when the victim is targeted because of age or disability. Mistreatment includes: physical abuse, sexual abuse, neglect, and verbal/emotional/psychological abuse, and financial exploitation. Elder mistreatment is unfortunately common, affecting 5–10% of community dwelling older adults each year (Lachs & Pillemer, 2015). Nursing home residents are at an even higher risk (Rosen et al., 2018). Older adults who experience mistreatment have a much higher mortality than other older adults, and mistreatment has been linked to poor medical outcomes including depression and worsening of chronic medical conditions (Lachs & Pillemer, 2015).

Unfortunately, elder mistreatment is infrequently detected, with research suggesting that only 1 in 24 cases is identified and reported to the authorities (Rosen et al., 2018). A health care encounter is an important and often underutilized opportunity to identify elder abuse or neglect, which may otherwise remain undiscovered. For many older adults, assessment by health care providers is their only contact outside the family. Despite this, limited existing literature suggests that elder mistreatment is very infrequently identified, diagnosed, or documented (Evans et al. 2017).

The International Classification of Diseases (ICD) is a list of codes corresponding to medical diagnoses that is published and periodically updated by the World Health Organization and is used worldwide to ensure standardized reporting on morbidity and mortality and for reimbursement systems. In the US, ICD code(s) are assigned to all health care encounters by health care providers or professional coders to indicate the medical conditions evaluated or treated during the encounter. These codes are used to determine appropriate costs and reimbursement for billing and insurance claims. Examining ICD codes within administrative insurance databases has been identified by researchers and policymakers as a uniquely powerful strategy to explore and improve understanding of patterns of health care utilization as well as diagnosis and treatment of individual diseases in the United States given the country’s fragmented health care system. Claims data offers an opportunity to examine the care provided by millions of health care providers to tens of millions of patients and, as a result, is widely used for surveillance and research for a broad range of medical conditions. Using this approach to identify and examine cases of elder mistreatment may yield results critical for the field to inform future research and surveillance practices.

Concerns exist, however, about how frequently and reliably these ICD codes are actually used. It is possible that health care providers may recognize abuse and even act on their elder mistreatment concerns but prefer to not code this diagnosis without an in-depth investigation or avoid documenting given wariness of the legal or patient safety ramifications of doing so. Also, providers may choose to code immediate medical issues, such as sepsis or fractured facial bones, rather than the underlying cause, which may be neglect or physical abuse.

In the area of child abuse, researchers have explored insurance claims as a potential source of information for case finding, to characterize patterns of health care utilization, to compare to other data sources, and to estimate cost. Findings from these studies have been useful, though they have concluded that use of ICD codes alone underestimates the phenomenon for similar reasons (Raghavan et al., 2015; Scott et al., 2009). The exploration of ICD codes for elder mistreatment has been much more limited to date, though ICD-9 (Ninth Revision) codes have been used previously for elder abuse case finding during ED visits (Evans et al., 2017) and hospitalizations (Rovi et al, 2009)

Notably, the transition to ICD-10, in October 2015, added 14 additional codes for suspected rather than confirmed elder mistreatment. This was due partially to advocacy from child abuse researchers and public health professionals and was intended to address issues with cases not being coded. Advocates believed that adding “possible abuse” codes would allow recording of potential mistreatment or ongoing investigation even where a definitive determination had not yet occurred (Raghavan et al., 2015; Scott et al., 2009). They suggested the addition of these codes might also allay the concerns of health professionals who believed definitively coding mistreatment as outside their scope of practice or who thought doing so might create legal issues for them or their patients (Raghavan et al., 2015; Scott et al., 2009). The impact and pattern of initial use of these codes in elder mistreatment has not been described, and describing it is important to inform future elder abuse research and surveillance strategies. We hypothesized that the addition of the “possible abuse” codes with the ICD-9 to ICD-10 transition would substantially increase the frequency of elder mistreatment diagnostic coding. Our goal was also to describe the frequency of diagnosis of different types of elder mistreatment in multiple health care settings and to examine differences in diagnosis by Medicare Advantage (MA) vs. private insurance status.

In this manuscript, we use a large database of patients in the United States to describe the diagnosis of elder mistreatment over 2011–2017, focusing on the impact of the ICD-9 to ICD-10 transition.

Methods

We examined elder mistreatment diagnosis in health care settings using data from the Health Care Cost Institute (HCCI) insurance claims database. The HCCI database contains detailed claims data for approximately 50 million individuals per year insured with Aetna, Humana, and UnitedHealthcare (including MA enrollees) in all 50 states and the District of Columbia. These insurers represent approximately 30% of all privately insured adults in the US. MA enrollees in the HCCI database accounted for close to 50% of all MA enrollees nationwide during the study period.

We focused on patients aged ≥65 years who were enrolled in an MA or private insurance plan for at least one month during 2011–2017. HCCI has developed and applied algorithms to link records pertaining to the same individual patient over time and across different plans to avoid counting patients multiple times. Similar to previous studies, we included in our analysis multiple ICD codes which identify different types of elder mistreatment (Evans et al., 2017). We categorized codes into those indicating: physical abuse; neglect or abandonment; sexual abuse; psychological or emotional abuse; and abuse, neglect, and maltreatment without further specification. Notably, none of these codes are specific to older adults. As the transition from ICD-9 to ICD-10 occurred within the period of our study (transition started in October 2015), both were included. For this study, we grouped codes both within and between ICD-9 and ICD-10 classification systems that represented the same type of mistreatment. This grouping was based on a consensus-building process involving several of the authors (TR, LKM, VML, MSL), all health care providers with extensive elder abuse expertise. To ensure appropriate grouping of codes between ICD-9 and ICD-10, we also consulted multiple available crosswalks, which map codes in the two versions of ICD that correspond to the same diagnosis to facilitate conversion. Details of the grouped codes, including the ICD description of each, are included as Supplementary Material. For each patient, we identified the first elder mistreatment diagnosis in our data (if any) defined as any one elder mistreatment ICD diagnosis code on insurance claims for healthcare encounters in any setting (e.g., inpatient, outpatient, Emergency Department [ED]). Notably, we report here on the number of patients receiving any elder mistreatment diagnosis (either a diagnosis indicating a specific type of mistreatment or a diagnosis suggesting abuse, neglect, and maltreatment without further specification) as well as: the number receiving a diagnosis of a specific type of mistreatment and the number receiving diagnoses of multiple types.

We examined the use of “suspected” codes in ICD-9 and ICD-10 in comparison to codes indicating definitive diagnosis. A single code existed in ICD-9 to indicate clinician suspicion of elder mistreatment: “observation and evaluation for suspected abuse and neglect” [V71.81]. In ICD-10, codes for 14 types of “suspected” mistreatment were added relevant to older adults, including “adult physical abuse, suspected” [T76.11] and 2 types of “alleged” mistreatment including “encounter for examination and observation following alleged physical abuse” [Z04.7].

We evaluated time trends over 2011–17 in quarterly population rates of elder mistreatment diagnosis by type of mistreatment diagnosis, sex, age category (65–74, 75–84, 85+), and by health care setting. When calculating rates, we used as the denominator patients enrolled for at least 1 month in a given calendar quarter.

We estimated linear regressions to assess time trends and statistical significance. To test whether the trend lines shifted and/or changed slopes from before to after the ICD-9 to ICD-10 transition, we estimated segmented regressions.

We also examined the type of health care settings in which the initial elder abuse diagnosis was made. For individuals whose claims indicated services received from multiple types of providers or in multiple settings at the time of their first diagnosis, we adopted the following hierarchy in determining the setting surrounding the diagnosis: 1, highest: hospital, 2: ED, 3: hospital outpatient, 4: physician office, 5: ambulance, 6: other. This hierarchy was established by the authors based on their knowledge of clinical assessment of elder mistreatment as well as insurance claims data.

We reported on patients with private insurance as well as MA patients because we discovered that a large number of adults aged ≥65 within the HCCI database were actually on private insurance. Given the potential for differences between these populations, we reported on them separately. We compared results for private insurance vs. MA patients, but we did not report tests of significance for these comparisons because, given large numbers, even small differences reached statistical significance.

This analysis was approved by the Weill Cornell Medicine Institutional Review Board, protocol number 1807019417.

Results

From 2011–17, 8,127 patients (0.051% of all 17,510,438 patients aged ≥65) received a diagnosis code of any elder mistreatment, including codes for “suspected” mistreatment. We compared “suspected” vs. “confirmed” codes, finding that, in ICD-9, 9.7% of cases were coded as “suspected” using the single available code. The use of “suspected” codes in ICD-10 for each type of abuse from the introduction of the ICD-10 in the 4^th quarter of 2015 until the end of 2017 is shown in Table 1. Notably, “suspected” codes were used in 45.3% of cases overall, most commonly in neglect or abandonment (53.8%) and least commonly in psychological or emotional abuse (21.8%).

Table 1.

Comparison of ICD-10 Coding of Suspected vs. Confirmed Elder Mistreatment in Insurance Claims, 4^th Quarter 2015 (when ICD-10 introduced) – 4^th Quarter 2017

Mistreatment Type	Suspected (%)	Confirmed (%)
Physical abuse	50.2%	49.8%
Neglect or abandonment	53.8%	46.2%
Sexual abuse	48.4%	51.6%
Psychological or emotional abuse	21.8%	78.2%
Abuse, neglect, and maltreatment without further specification	38.0%	62.0%
Any elder abuse	45.3%	54.7%

The time trend in rates of diagnosis of different types of elder mistreatment are shown graphically for MA and private insurance patients in Figure 1. The transition from ICD-9 to ICD-10, shown in the figure with a vertical line, appears to be associated with an increase in the rate of diagnosis.

Figure 1:

Trends in Diagnosis of Different Types of Elder Mistreatment among Medicare Advantage and Private Insurance Patients Aged ≥65, 2011–17

Included among patients receiving any elder mistreatment diagnosis code were 6,304 patients with MA (0.058% of all 10,908,425 MA patients) and 1,823 patients with private insurance (0.026% of all 6,982,317 private insurance patients). Types of mistreatment coded and patient sex are shown in Table 2. The most common type of mistreatment diagnosed in both MA and private insurance patients was physical abuse followed by neglect or abandonment. Women much more commonly than men received a diagnosis each type of elder mistreatment among both MA and private insurance patients. Figure 2 shows the trend in diagnosis by sex, highlighting that the overall increase during the study period from ICD-9 to ICD to ICD-10 is driven by the increase in diagnoses of women. Figure 3 shows the rates of diagnosis by age category. Rates of elder mistreatment for the entire study period for both MA and private insurance patients were highest among patients aged 65–74 years, followed by those aged 75–84, with the lowest rate among patients aged 85 years or older. Health care settings in which these diagnoses were made is shown in Table 3. The most common settings for diagnosis coding among MA patients were the ED, physician offices, and hospitals. For patients with private insurance, physician offices, hospital outpatient, and hospital settings were most common, with the ED playing a much smaller role. Time trends in rates of diagnosis coding within different health care settings are shown in Figure 4. Notably, the largest increase in diagnosis rate during the study period was the increase among MA patients in EDs, which increased 2.53 fold from 2011–2017.

Table 2.

Frequency of Types of Mistreatment Identified among Medicare Advantage and Private Insurance Patients Aged ≥65, 2011–17

			Female		Male
Mistreatment Type	n	% of all patients	n	%	n	%
Medicare Advantage
Physical abuse	1,877	0.017%	1,343	0.022%	534	0.011%
Neglect or abandonment	1,427	0.013%	957	0.016%	470	0.010%
Sexual abuse	550	0.005%	482	0.008%	68	0.001%
Psychological or emotional abuse	492	0.005%	412	0.007%	80	0.002%
Abuse, neglect, and maltreatment without further specification	2,227	0.020%	1,642	0.027%	585	0.012%
Multiple types of mistreatment	84	0.001%	67	0.001%	17	0.000%
Any elder abuse	6,304	0.058%	4,624	0.075%	1,680	0.035%
Private insurance
Physical abuse	539	0.008%	381	0.011%	158	0.005%
Neglect or abandonment	370	0.005%	218	0.006%	152	0.004%
Sexual abuse	146	0.002%	123	0.003%	23	0.001%
Psychological or emotional abuse	142	0.002%	110	0.003%	32	0.001%
Abuse, neglect, and maltreatment without further specification	697	0.010%	472	0.013%	225	0.007%
Multiple types of mistreatment	21	0.000%	13	0.000%	8	0.000%
Any elder abuse	1,823	0.026%	1,257	0.035%	566	0.017%

Figure 2:

Trends in Diagnosis of Elder Mistreatment by Sex among Medicare Advantage and Private Insurance Patients

Figure 3:

Trends in Diagnosis of Elder Mistreatment by Age Group among Medicare Advantage and Private Insurance Patients

Table 3.

Frequency of Types of Mistreatment Identified by Health Care Setting among Medicare Advantage and Private Insurance Patients Aged ≥65, 2011–17

Mistreatment Type	Hospital (% of cases identified in all settings)	ED (%)	Hospital Outpatient (%)	Ambulance (%)	Physician Office (%)	Other (%)
Medicare Advantage
Physical abuse	19.8%	39.2%	6.9%	5.1%	19.4%	9.6%
Neglect or abandonment	40.7%	26.4%	8.1%	3.3%	10.2%	11.3%
Sexual abuse	7.3%	47.8%	4.5%	15.6%	14.5%	10.2%
Psychological or emotional abuse	11.2%	13.6%	11.6%	0.6%	50.8%	12.2%
Abuse, neglect, and maltreatment without further specification	19.3%	17.5%	11.7%	13.9%	29.3%	8.3%
Multiple types of mistreatment	26.2%	46.4%	4.8%	2.4%	17.9%	2.4%
Any elder abuse	22.1%	26.9%	9.2%	8.5%	23.3%	10.1%
Private Insurance
Physical abuse	22.6%	17.1%	21.9%	1.9%	27.3%	9.3%
Neglect or abandonment	51.1%	10.0%	17.6%	2.2%	11.9%	7.3%
Sexual abuse	3.4%	24.0%	37.0%	4.1%	21.9%	9.6%
Psychological or emotional abuse	19.0%	6.3%	12.0%	0.0%	58.5%	4.2%
Abuse, neglect, and maltreatment without further specification	16.9%	6.0%	29.3%	11.8%	29.1%	6.9%
Multiple types of mistreatment	23.8%	23.8%	14.3%	4.8%	28.6%	4.8%
Any elder abuse	24.0%	10.8%	24.1%	5.8%	27.4%	7.8%

Figure 4:

Trends in Diagnosis of Elder Mistreatment in Different Health Care Settings among Medicare Advantage and Private Insurance Patients

Discussion

The findings we report here represent, to our knowledge, the first use of insurance claims data to examine rates of elder mistreatment diagnosis coding. Our results describe the current low use of ICD codes by health care providers in the US to diagnose elder mistreatment. Our findings will also inform decisions about the potential for insurance claims data to be useful for elder mistreatment case findings and analysis in future research and surveillance.

We found that rates of diagnostic coding of elder abuse were very low. The transition from ICD-9 to ICD-10 was associated with a small but substantial increase in rate of diagnosis, with ICD-10 “suspected” codes used frequently (Table 1), accounting for nearly half of all cases during the ICD-10 portion of the study period. Many of the cases coded as “suspected” elder mistreatment may not have been previously documented by ICD-9 codes due to the absence of a “suspected” category. Hunter and colleagues found in examining use of ICD-10 codes for child abuse between 2016–2018, 58% were suspected rather than confirmed maltreatment, suggesting widespread use of this newly available option (Hunter et al., 2020). Though the addition of “suspected” codes may have modestly increased overall coding of elder mistreatment, it may also create issues with surveillance and research in differentiating between suspected and confirmed cases. Many cases that would have been coded as confirmed in ICD-9 may be coded as suspected in ICD-10, with the codes likely not applied in a uniform way. Thus, tracking the rate of elder mistreatment and conducting research about the phenomenon will require determining whether to include suspected cases and, if so, whether to combine them with confirmed cases or analyze them separately. Strategies will need to be developed by public health professionals and researchers to address these issues. For surveillance, strategies may include tracking suspected cases to explore whether and when the diagnosis is confirmed. Researchers investigating elder abuse in the future using ICD codes may conduct sensitivity analyses excluding and including suspected cases to evaluate the impact on their findings.

Overall, we found that all types of elder mistreatment were very infrequently diagnosed, with only 0.051% of all patients aged ≥65 (or 1 in 1,960 patients) receiving any mistreatment-related diagnosis. Very low rates were present for both MA and private insurance patients (Table 2). This supports existing research and clearly demonstrates the limitations of using ICD coding alone to identify cases of elder mistreatment for surveillance or research. Rovi and colleagues (2009) examined the Nationwide Inpatient Sample, finding that <0.02% of hospitalizations were coded with an elder mistreatment diagnosis. Evans and colleagues (2017) examined the National Emergency Department Sample and found that elder mistreatment was diagnosed in only 0.013% of ED visits. This low rate of diagnosis is despite consistent findings from population-based studies suggesting that the phenomenon is common. It is possible that elder mistreatment episodes may not lead health care utilization. It is likely that some cases of elder mistreatment were identified and diagnosed but not coded. This is similar to other social determinants of health, including food insecurity and unstable housing, which are also infrequently coded by health care providers. Additionally, providers may hesitate to code due to fear of stigmatizing the older adult or concern that the patient or abuser will see mistreatment suspicion documented in the medical record, undermining the therapeutic alliance or creating danger for reprisal. This issue is likely amplified with increased focus on “open notes,” an initiative where all clinical documentation is immediately released to patients (Walker et al., 2015). Our research also reinforces that health care providers seldom identify elder mistreatment, strongly suggesting that opportunities to do so are usually missed. Screening for elder mistreatment in health care settings (Rosen, Platts-Mills, & Fulmer, 2020) as well as additional education for health care providers (Alt et al., 2011; Ejaz et al., 2020; Rosen et al., 2019) may be helpful.

While all types of mistreatment were diagnosed at rates much lower than would be anticipated based on the population prevalence, we noted that some types were diagnosed more commonly than others. Population-based studies suggest that elder physical abuse (0.2–2.1% annually) is much less common than neglect (5.1–5.4%) or verbal/emotional/psychological abuse (4.6–12.9%), and sexual abuse (0.3–0.6%) is even less common (Acierno et al., 2010; Amstadter et al., 2011; Laumann, Leitsch, & Waite, 2008). Recent research examining calls to a US nationwide elder abuse resource line also found that neglect and verbal/emotional/psychological abuse were reported more frequently (Weissberger et al., 2020). Physical abuse and sexual abuse were more commonly diagnosed among MA and private insurance patients than other types, particularly in the ED setting. This finding may be because these types of mistreatment are more likely to lead to emergent health care encounters (e.g. a facial laceration due to physical abuse) or because they are easier to recognize and confidently diagnose. Neglect may be harder to identify, as many of the findings overlap with chronic illness and aging. Verbal/emotional/psychological abuse may be less commonly diagnosed because it may be perceived to have fewer significant medical consequences. As a result, health care providers may either not assess for it or believe it is not important enough to code even if identified/suspected.

Notably, the existing literature suggests the victims of one type of elder mistreatment are at high risk for others and that many victims suffer from multiple types of mistreatment concurrently (Lachs & Pillemer, 2015). Patients in the cohorts we examine here, however, were very seldom diagnosed with multiple mistreatment types at the same time. It is possible that health care providers chose to code only one type of mistreatment (perhaps the most serious) or attempted to choose a code/code(s) that they felt most accurately represented the entire clinical picture. Nevertheless, educating health care providers that identifying one type of elder mistreatment should lead to careful assessment for others may be useful.

The diagnosis of elder mistreatment, particularly physical abuse and neglect or abandonment, substantially increased during the study period (Figure 1). This suggests that health care providers were more frequently recognizing and diagnosing elder mistreatment, potentially due to increased education and awareness, in addition to the impact of the ICD-9 to ICD-10 transition. An important part of this increase was growth in ED diagnosis of elder mistreatment, particularly among MA patients. Recent efforts to increase recognition of elder mistreatment in this setting, including focus on geriatric emergency care (Rosen, Hargarten, Flomenbaum, & Platts-Mills, 2016; Rosen et al., 2020; Rosen, Stern, Elman, & Mulcare, 2018) may have been impactful. Geriatric ED guidelines published in 2014 advocated for additional education and quality improvement focused on elder abuse and neglect (“Geriatric emergency department guidelines,” 2014). Also, a review article on elder abuse (Lachs & Pillemer, 2015) was included in the American Board of Emergency Medicine’s 2018 Lifelong Learning Self-Assessment Maintenance of Certification Curriculum (American Board of Emergency Medicine., 2017).

Previous population-based research examining whether sex or age increases risk of elder mistreatment has been inconsistent (Dong, 2015; Pillemer et al., 2016). We found that women and patients from younger-old age groups (aged 65–74 and 75–84) more commonly received a diagnosis of elder mistreatment from health care providers (Table 2, Figures 2,3). This may suggest that mistreatment has a higher prevalence in these groups. It is also possible that younger-old patients and women are more likely to report elder mistreatment to health care providers or that health care providers more commonly consider the potential for elder mistreatment in these patients.

Our findings revealed important differences between the MA and privately insured patients. Rate of diagnosis was more than twice as high among MA patients as among private insurance patients (Table 2). Neglect was much more commonly diagnosed, and an elder mistreatment diagnosis was made more commonly in the ED, among MA vs. private insurance patients. These differences were likely driven by the differences in these populations. It is possible that older adults with MA are more commonly mistreated severely enough to require care in the ED. Also, for privately insured older adults, physician office visits are also less likely to be jammed with competing priorities, so mistreatment may more likely surface during these encounters. Additional research is needed to further explore and characterize differences in elder mistreatment between these two populations.

This research has several limitations. We report here on trends from 2017 and earlier as more recent data were not available to our research team. Changes may have occurred subsequently as providers became more familiar and comfortable with ICD-10. We believe, however, that our findings are informative and generate hypotheses that may be evaluated on subsequent data when it becomes available. We are unable to use HCCI data to assess whether confirmed or suspected mistreatment was reported to Adult Protective Services (APS) or another entity for additional investigation, as this information is not included in medical insurance claims. As health care providers are mandatory reporters in most US states (Geiderman & Marco, 2020), it is likely that the majority of these cases were reported, , but we are unable to be certain of this. Also, it would be interesting to know whether APS opened a case based on the report from a health care provider and the results of their subsequent investigation. Further, health care providers may have made a referral to other aging services professionals to intervene (Dauenhauer et al., 2019). Ultimately, it would be useful to explore whether health care provider recognition of elder mistreatment as well as reporting and any subsequent intervention reduced the risk of re-victimization (Burnes et al., 2020). Our research presumes that all diagnoses of elder neglect represent mistreatment by another person rather than self-neglect. We recognize that self-neglect is a common phenomenon that accounts for more repeat referrals to Adult Protective Services than elder mistreatment by another person (Rowan et al., 2020). Additionally, we are not aware of an ICD-9 or ICD-10 code for self-neglect. While most health providers and professional coders likely understand that the adult neglect codes indicate neglect by another person, it is possible that some providers may use them in self-neglect cases, thus introducing misclassification in our results.

The potential for using claims data to improve elder mistreatment identification within health care settings has been described (Rosen, Zhang, et al., 2019). Our work represents a critical preliminary step in assessing this potential for researchers, policymakers, and health care providers by describing the frequency of diagnosis, characteristics of victims and health care settings in which diagnosis is made, and the impact of the ICD-9 to ICD-10 transition.

Notably, the addition of suspected codes in the transition from ICD-9 to ICD-10 coincided with increasing diagnosis of elder mistreatment. This trend is encouraging, despite the fact the it may create new challenges for research and surveillance in differentiating between suspected and confirmed cases. This increase in diagnosis may also be partially driven by greater recognition of this phenomenon. Overall, though, our work confirms existing findings from smaller studies that elder mistreatment is likely dramatically under-recognized and under-diagnosed in health care settings. Our findings about patterns of diagnosis of individual types of elder mistreatment and in different health care settings suggest opportunities for additional exploration to improve our understanding. Future research is needed to further explore health care utilization patterns among elder mistreatment victims before and after diagnosis to improve understanding of the burden of this syndrome.

Creating a conceptual model drawn from the literature on risk factors of elder mistreatment (e.g,. age, sex, diagnoses of dementia, depression, substance abuse) may allow future studies to examine predictors of abuse diagnosis (any type) or specific types of abuse in a healthcare setting. Ultimately, research strategies are critical to explore the much larger group of victims whose mistreatment goes un-diagnosed.