A Retrospective Cohort Study of Factors that Affect Healing in Long-term Care Residents with Chronic Wounds
Chronic ulcers such as pressure, ischemic, and venous ulcers are common in long-term care (LTC) and frequently do not heal. A retrospective medical records review of all LTC residents referred to a wound consultative service between April 1999 and January 2007 was conducted to assess predictors of 6-month healing outcome.
Variables abstracted and analyzed included wound, resident demographic, and laboratory values at diagnosis and comorbid medical illnesses. The average age of study participants (n = 397) was 78.1 years (± 11), 47% were men, 48% had more than one wound, and the most common wound diagnosis was pressure ulcer (n = 163). After 6 months, 66% of ulcers were not healed. The odds ratio for nonhealing was significantly higher in residents who had more wounds, a larger wound area, diabetes mellitus, or peripheral vascular disease and lower in residents with increased age and hemoglobin values and/or a history of stroke, depression, dementia, degenerative arthritis, peripheral neuropathy, and falls. After adjustment in the multivariate model, only the number of wounds and hemoglobin level remained significant predictors of healing status. A higher number of chronic ulcers and lower hemoglobin counts increased the risk of nonhealing after 6 months of care. Including these variables in LTC resident assessments may help clinicians ascertain expected outcomes of care.
Healthcare providers face many challenges in treating chronic ulcers. These ulcers commonly occur in long-term care (LTC) facilities and include pressure, ischemic, venous, and ulcers of mixed etiology. The prevalence of pressure ulcers among LTC residents ranges from 11% to 15%1-3; the prevalence of venous ulcers is estimated to be 2.5%.4 Chronic ulcers decrease LTC resident quality of life and increase morbidity and mortality as well as overall healthcare costs.5 In adults >75 years of age with a pressure ulcer, the 12-month mortality rate is 29% and relative risk of mortality in the ulcer group is 1.92 (95% confidence interval [CI], 1.52 to 2.43) compared to persons of similar age who are ulcer-free.6
Mortality is only one concern; many patients want to know about their prognosis for healing. While the effect of some wound characteristics — eg, pressure ulcer severity — are known, 7 the effect of comorbid illness remains controversial and information from previous retrospective cohort studies7-9 is conflicting.
The primary aim of chronic ulcer management is to prevent occurrence and, if ulcers do develop, to facilitate complete healing. Previous retrospective cohort studies8,10 that have evaluated healing characteristics only for pressure ulcers in LTC found healing generally improves with a moist wound environment and that larger wounds take longer to heal than smaller wounds. Although these studies provide important information on characteristics associated with healing, many ulcers in the studies were not easily classified and represented mixed etiologies.11 To help fill deficits in the knowledge base for predictive factors of healing in all ulcer types, a retrospective cohort study was conducted to identify wound and patient characteristics that can predict complete ulcer healing in residents of LTC facilities.
Methods and Procedures
A retrospective cohort study of ulcer healing was conducted among adult residents with a physician-confirmed diagnosis of a pressure or other chronic ulcer (ischemic, venous, neuropathic, mixed etiology) living in Olmsted County, Minnesota LTC facilities. The Mayo Clinic Institutional Review Board (IRB) reviewed and approved the protocol. All aspects of this research were conducted in accordance with the principles of the Declaration of Helsinki12 and state law. 13
Participants. Investigators retrieved data from all the records of consecutive adults ages 18 and older with an incident diagnosis of pressure, ischemic, venous, neuropathic, or mixed ulcer evaluated between April 1, 1999 and January 1, 2007 by the LTC wound consultative service, which comprises midlevel healthcare providers (medical doctors and physicians assistants with expertise in wound care) and consultative physician staff. The LTC wound care service only sees wounds following referral from primary care providers at the eight LTC facilities. Each facility has 24-hour nursing support and participates in the state survey process. The date of ulcer diagnosis was defined as the initial wound care visit date.
Inclusion and exclusion criteria. Participants had to be residents of one of the eight skilled nursing facilities affiliated with Mayo Clinic and have completed at least one wound-related follow-up visit with a healthcare provider. Patients were excluded if they had quadriplegia or paraplegia, were enrolled in hospice, and/or did not provide informed consent.
End points. The primary outcome was healed or not healed. Complete healing was defined as 100% closure of the wound or, if more than one wound was present, closure of the primary (index) wound. The healthcare providers established and documented healing status during follow-up visits based on clinical characteristics of the ulcer. An ulcer was considered not healed if complete healing was not documented within 6 months of the initial medical visit or if the resident was lost to follow-up or died.
Study definitions of ulcer types. The operational criteria for a pressure ulcer included a new area of ulceration over a bony prominence such as the sacrum or ischium. Ischemic ulcers were defined as an ulcer that occurred in persons with arterial insufficiency normally involving a limb. The healthcare provider diagnosed ischemic ulcers based on the clinical findings at the time of the visit (with or without subsequent arterial testing).11,14 Venous ulcers were lower leg wounds in residents with venous insufficiency or edema15; foot ulcers in an area subject to pressure in residents with peripheral neuropathy were classified as neuropathic ulcers. 16,17 An ulcer was characterized as being of mixed etiology if two or more overlapping components (eg, ischemic disease in a neuropathic ulcer) were diagnosed by the clinical provider.
Independent variables. The primary predictive variables included wound characteristics — ie, ulcer area (cm2), number of ulcers, and ulcer type. Wound area was the calculated product of measured length and width in cm. Number of ulcers was the counted number of ulcers on initial presentation.
Demographic data included date of birth (age in years), gender, nursing home location, and calculated body mass index (BMI) in kg/m2 at the time of diagnosis. Laboratory assays were performed up to 1 year before the initial wound consultation and included most recent hemoglobin, white blood cell count, and creatinine.
All medical conditions present per physician diagnoses at the time of initial consultation were noted and included coronary artery disease (CAD), congestive heart failure (CHF), peripheral vascular disease (PVD), venous insufficiency, and renal insufficiency. Other prevalent comorbid conditions documented in the medical record included cancer (nonskin), diabetes, depression, dementia, blindness, rheumatoid arthritis, degenerative joint disease (DJD), peripheral neuropathy, and previous fall.
Data collection and storage. Blinded abstractors obtained data directly from the wound registry, medical records, and administrative data within the Mayo Clinic Rochester health records system. The Mayo Clinic maintains all electronic medical record (EMR) information within one longitudinal records linkage system18 that includes documentation of nursing home care. Blinded information technology personnel collected laboratory information and comorbid health conditions via abstraction from the EMR.
Data analysis. The study abstractor entered all variables into Microsoft Excel (version 2003, Microsoft, Redmond, WA). The primary data were stored in an electronic system and continuously saved (daily) and backed up by optical disk. All statistical analyses were performed utilizing SAS 9.1.3 software (SAS Institute Inc., Cary, NC). A univariate analysis was used to compare baseline patient demographics between participants whose ulcers had or had not healed within the study period. For risk factors, the mean ± standard deviation (SD) for continuous and frequencies and percentages for categorical variables were compared by univariately fitting a logistic regression analysis. Logistic regression models were used to obtain odds ratios (ORs) to predict healing status and 95% confidence intervals around the ORs. Variables meeting a P value <0.1 in univariate analysis were included in preliminary multivariate models. Multivariate logistic regression was used to derive the final model using a backward elimination approach that was confirmed by a stepwise approach. All P values <0.05 were considered statistically significant. After model development, the investigators tested for interactions between the final predictors in the model.
Of the 440 residents evaluated by the wound consultation service for a chronic ulcer during the study period, 411 gave authorization for their medical records to be examined for research purposes. Fourteen residents had either quadriplegia or paraplegia and were excluded, yielding a final study population of 397. At the time of the initial wound visit, the average age of participants was 78.1 years ± 11.25, 47% were male, average BMI was 30.2 kg/m2 ± 14.7, and 48% had more than one ulcer. Of the five types of ulcers, pressure ulcers (n = 163) and ischemic ulcers (n = 96) were most common; 107 residents had ulcers of mixed etiology.
The ulcers of 262 study participants (66%) did not completely heal within the study period; of the 397 patients, 92 (24%) died within 6 months of initial evaluation (see Table 1). The average age in the nonhealed group was lower than the healed group (76.71 years ± 11.8 versus 80.7 years ± 9.6, respectively; P <0.001). One third (133, 33%) of participants resided in the nursing home that specializes in wound care. Of the 117 participants with PVD, 91 did not heal (OR = 2.25 [95% CI 1.37–3.69]). The difference in average baseline ulcer area between the healed and the nonhealed group was small but statistically significant (OR = 1.02 [95% CI 1.01–1.04]) and the average number of ulcers per resident in the nonhealed group was also higher than the average number of ulcers in the healed group (OR = 1.42 [95% CI 1.13–1.78]).
The 13 significant predictors (P <0.10) of healing outcome were included in the stepwise analysis. In this model, the number of ulcers per resident continued to be associated with an increased risk of nonhealing but average ulcer area was no longer significant. Increasing age trended toward a decreased risk of nonhealing and higher hemoglobin level decreased the risk of nonhealing (see Table 2). All remaining factors in the vascular and comorbid illness domains were not significant after adjustment. After calculation of the final model, no interactions were noted between the variables in the final model.
It is well recognized that residents of LTC facilities are at high risk for developing chronic ulcers that may or may not heal.19 The results of this study indicate that having multiple ulcers (as was the case in 48% of the study population) decreased the likelihood of healing after 6 months — unadjusted OR = 1.42 (95% CI 1.13–1.78). Although not previously reported, this observation is biologically plausible.
The presence of multiple ulcers may indicate a poor overall health status and the presence of underlying physical conditions such as altered nutrition, decreased functional mobility, or reduced skin perfusion. Larger ulcer area was associated with nonhealing in the unadjusted analysis (OR 1.05 [95% CI 1.10–1.04]) but not in the multivariate analysis. In an exploratory study of elderly community and LTC residents (N = 82), 8 larger pressure ulcers completely healed less frequently than smaller ulcers. Large ulcers (>20 cm2 ) healed completely 15% of the time in 6 months compared to 45.9% of ulcers <5 cm. 20 In a prospective cohort study21 of neuropathic diabetic foot ulcers, larger ulcer size was associated with longer healing times. These observations confirm the importance of rigorous wound measurement and evaluation.
Hemoglobin also was significantly associated with wound healing — higher hemoglobin increased the odds of healing. The OR for hemoglobin model was 0.78 (95% CI 0.66–0.92) in the unadjusted and 0.70 (95% CI 0.56–0.87) in the multivariate model. Hemoglobin concentration has been associated with different outcomes in many pressure ulcer wound studies22,23 and has been reported to influence incidence, nonhealing, and mortality; lower hemoglobin levels were associated with pressure ulcer development in postoperative patients. 24 Hemoglobin also is a component of a new instrument used to predict pressure ulcer development. 25 Thus, obtaining hemoglobin levels may have potential value; however, the safety and effectiveness of treating anemia is still under investigation. The current finding, noting the association between higher hemoglobin concentration and lower nonhealing rates, may encourage further research regarding increasing hemoglobin levels in LTC residents with chronic ulcers. Further study assessing relative change from baseline hemoglobin and longitudinal assessment of hemoglobin over time (unable to be studied in the current investigation) is warranted.
Age and male gender also were factors in nonhealing in this study. After adjustment, increasing age trended toward decreasing nonhealing status but not significantly. The rationale for why older patients were more likely to heal is not certain. In the Jones and Fennie study, 8 the average age of the healed population was slightly but not significantly higher than the nonhealed population. In their study of 125 pressure ulcer patients, Domini et al9 found no differences in age between the population that healed or did not heal. In the nonadjusted analysis of the current study data, male gender was a risk factor for nonhealing. These findings may reflect a higher burden of comorbid illnesses in the male gender, but gender has not played a role in other studies of pressure ulcer healing. 8,9 In a recent British prospective cohort study26 of persons with venous ulcers, gender had no impact on healing.
Body mass index was not associated with 6-month nonhealing status. Although nutritional status has been associated with pressure ulcer healing in some studies, 9 specific and reliable data on nutritional status were not available for the current investigation and in the final analysis BMI was not associated with healing status.
Previous stroke and the presence of PVD had a significant influence on wound healing in univariant modeling, while other vascular conditions (cardiac disease, congestive heart failure, renal insufficiency, and venous insufficiency) were not found to be significant in the model. Peripheral vascular disease was associated with increased risk of nonhealing; this is plausible considering that PVD has been found to be a leading cause of lower extremity ulceration that increases with age27 and would be expected to increase nonhealing rate. In cohort studies, 28 up to 83% of patients with ischemic ulcers did not heal and often required amputation. After adjustment, PVD was not found to be a significant factor for nonhealing status in this study, which may reflect the small number of ischemic wounds in the current study population. A history of stroke increased the odds of healing but after adjustment this variable was not significant. The lack of association between the other vascular risk factors evaluated and nonhealing has been observed in other retrospective cohort studies of pressure ulcers, with vascular disease (except CAD) having no significant effect on pressure ulcer healing. 8
Although other comorbid health problems influenced nonhealing status on univariable analysis, none were significant after adjustment. Not surprisingly, diabetes trended toward an increased risk of nonhealing on initial analysis — diabetes is a leading cause of nonhealing diabetic foot ulcers16 — but after adjustment in multivariate modeling, diabetes was no longer significant. Depression, dementia, degenerative arthritis, history of falls, and peripheral neuropathy also were included in the final model. Interestingly, the odds of nonhealing were lower in patients with these comorbid illnesses but not independent of age and other health conditions included in the multivariate analysis. Increased age in the healed group likely accounted for a large part of the protective effect seen on unadjusted analysis. The possibility of survivorship bias (in which those patients with illnesses who survive to the nursing home are different than those who did not) also could potentially explain some of the initial findings. Although these findings were not expected, the literature includes precedence for the lack of an association between comorbid health conditions and wound healing, Specifically, Jones8 found that a history of PVD, diabetes, cancer, arthritis, neuropathy, dementia, stroke, depression, sensory disorders, and renal disorders was not associated with improved healing, which would be consistent with current findings. However, Jones8 did find an association between coronary disease and healing rates that differs from current findings. In a study10 of 882 LTC residents with pressure ulcers, vascular disease, age, and diabetes did not influence pressure ulcer healing and dementia actually increased wound healing rates. In other studies,8 the presence of diabetes, cancer, dementia, or depression was not associated with 6-month pressure ulcer healing. The results of the current study confirm that the potential effect of comorbid health status on ulcer healing appears small.
The proportion of ulcers healed in the current study (34% in 6 months) was lower than that reported in other studies. In a pseudo-randomized pragmatic cluster trial28 of pressure and venous leg ulcers among LTC residents with a similar age range (83.0 years ± 9.1), a 52% healing rate (48% nonhealing rate) was observed. However, this study did not include ischemic or neuropathic ulcers. In Domini et al’s9 pressure ulcer study, 46.4% of ulcers were healed after 3 months — participant average age was 78.3 years ± 11 and subjects had an average of 3.9 ± 2 comorbid illnesses. These studies feature different populations and different wound types; the current study sample only included a referral population that may suffer from more severe, larger, or treatment-resistant ulcers compared to community samples. A large proportion (24%) of the subjects died within 6 months of initial wound evaluation, which also likely accounts for the higher nonhealing rate in the current study. An accurate expected healing rate in the LTC population with mixed etiology chronic ulcers, including ischemic wounds, has yet to be determined.
This study describes the relationship between wound and patient characteristics and 6-month healing status of chronic wounds. One limitation involves the retrospective cohort design and the potential for missing information. Although medical record information was fairly complete, some providers did not record certain relevant information, which could have lead to some bias in the results, decrease the power to detect differences, and present the potential for misclassification. The definition of documented healing, used in this study, and classifying all patients lost to follow-up as not healed ensured a conservative interpretation of the outcome. Because some wound characteristics (eg, size) were missing from some medical records, it is possible that small initial ulcers were not included. Clinicians seemed most concerned about healing Stage II and deeper ulcers. Also, the participants in this study were seen following referral to the wound care service and could be sicker, or have more recalcitrant wounds, than the general resident population in the LTC facility.29 Finally, electronic abstraction of comorbid illnesses from medical records could result in missing an underlying diagnosis or misclassification of a chronic wound. However, in previous studies30 that relied on electronic abstraction of Mayo records using electronic and recorded diagnosis, the accuracy has been >98%. The use of electronic diagnosis data is preferable when compared to reliance on claims data. 31
A 6-month retrospective cohort study of 397 LTC residents provides important new information about factors that influence nonhealing of chronic wounds. An increase in the number of wounds raised the risk of nonhealing 44%. Other wound characteristics (wound size) also predicted nonhealing in univariable modeling; however, in the full model, ulcer area was not significant. A higher hemoglobin count decreased the odds of nonhealing status in this population, while age, male gender, diabetes, stroke, PVD, depression, dementia, degenerative arthritis, falls, and peripheral neuropathy were significant on univariable testing only. The primary consideration for healthcare providers in LTC is a realistic expectation of healing. In this study, 66% of the ulcers had not healed in 6 months. In patients with multiple ulcers and a low hemoglobin count, the risk of nonhealing is even higher. Prospective clinical studies to assess the effect of treating lower hemoglobin counts on chronic wound healing appear warranted.
Dr. Takahashi is an Associate Professor of Medicine; Mr. Kiemele and Dr. Chandra are instructors; Mr. Cha is a statistician, Department of Biostatistics; and Dr. Targonski is an Associate Professor of Medicine, Mayo Clinic College of Medicine, Rochester, MN. Please address correspondence to: Paul Y. Takahashi, MD, Mayo Clinic College of Medicine, 200 First Street SW, Rochester, MN 55905; email: Takahashi.email@example.com.