The concept of dynamic frailty: an exploratory study of the trajectory to postoperative mortality

Abstract

Background

Frailty is a heightened vulnerability to stress due to decreased physical and mental abilities. Preoperative frailty has been associated with poorer outcomes. However, frailty is not static, and those patients who eventually die appear to become more frail. Our hypothesis is in-patient, postoperative changes in frailty after major operations predicts the trajectory to postoperative discharge alive or in-hospital mortality.

Study design

The accumulating deficit model of frailty was used. Data from the medical records of patients who have undergone major operations were used to determine the mFI preoperatively, postoperative day 1, and day before discharge or death. Of the 1063 patients who met inclusion criteria, 50 patients with in-hospital postoperative death and 50 patients discharged alive were randomly selected.

Results

Patients in the in-hospital mortality group had significantly greater median preoperative mFI scores than those in the discharged alive (0.178 vs. 0.115 p = 0.00009). This significant difference was present on postoperative day 1, while also increasing in margin (0.240 vs. 0.143, p < 0.00001). Median Pre-Post mFI differences were also significant between the two groups, with operations leading to in-hospital mortality experiencing a greater increase in mFI (0.06 vs. 0.01 p = 0.00019), and the day before death or discharge (0.276 vs. 0.014, p < 0.00001).

Conclusion

Preoperative mFI is a useful predictor of postoperative mortality. Moreover, worsening mFI score as early as day 1 and continued worsening scores throughout hospitalization are associated with a postoperative trajectory toward mortality. Recognition of worsening frailty may be helpful in identifying patients in need of early intervention.

Introduction

Frailty is the phenotypic state of physical and cognitive decline typically associated with older patients [1]. It has been associated with numerous adverse outcomes across a spectrum of health issues. Specifically, frailty has been associated with increased adverse postoperative outcomes across several specialties [2–6].

There are numerous ways that frailty can be measured. They may include some aspects of patient history, pre-existing health conditions, physical examination findings, laboratory abnormality, and specific physical and cognitive tests. One of the most tested methods is the “frailty index” based on the accumulating deficits model [7, 8]. This has the advantage of not requiring an in-person patient evaluation to determine the presence of the robust, pre-frail, or frail state. That is, the frailty index can be determined by evaluating medical records or any patient-specific data set. Iterations of the original frailty index are often referred to as a “modified” frailty index (mFI). These indices can be derived from the medical record based on the data available retrospectively. Of the original 70-item frailty index, indices with far fewer items, in some studies as few as five, have been proven predictive of adverse outcomes [6, 9–11].

Preoperative calculations of mFI’s are predictive of the occurrence of postoperative complications and, especially, postoperative mortality [6]. A recent study using a large database suggested that the accumulation of additional discharge diagnoses compared to admission diagnoses was associated with postoperative mortality [11]. This led to the concept of “dynamic frailty.” We define the concept of dynamic frailty as the accumulation of additional deficits that can occur after the patient has an operation. If these deficits increase to a certain level, the death will be the result. Therefore, the increase in postoperative frailty is the trajectory to postoperative mortality in these patients. Conversely, if the operation results in minimally increasing or even reducing the number of deficits, then this identifies the trajectory to recovery. Our purpose is to test the concept of dynamic frailty in a wide variety of operations as the mFI was shown to be predictive of mortality in nearly all types of operations [6]. We hypothesize that the mFI will increase (i.e., patients will become more frail) in patients who will eventually suffer in-hospital mortality compared to those who will be discharged living in a cohort of patients undergoing a wide variety of high-risk operations.

Methods

This study was approved by the Institutional Review Board of the University of South Florida Morsani College of Medicine.

Population

Data from the electronic medical records of patients 18 years or older who have undergone major operations from 2013 to 2022 at Tampa General Hospital were considered eligible for the study. Operations with a relatively high risk of mortality were chosen for the study in order to have enough patients suffering a postoperative death for analysis. These included open mitral/aortic valve replacement, pneumonectomy or pulmonary lobectomy, open abdominal aortic aneurysm repair, esophagectomy, proximal/total pancreatectomy, and partial/total colectomy/proctectomy. Patients were excluded if they were under 18 years old, involved in ongoing medical litigation, incarcerated, or did not undergo qualifying procedures in the specified timeframe. Patients who died during the index operation were the in-hospital mortality (IHM) group, while those discharge alive were the discharge living (DL) group. The DL group included patients discharged home or to another facility. As this was a pilot study to test the concept of dynamic frailty, a convenience sample of 50 patients were randomly selected from both the IHM and DL groups. For IHM, this was done by selecting every 6th patient record, and for DL, this was done by selecting every 14th patient record. The IHM patients were defined as the cases and the DL patients were the controls for a case-control study design.

Calculation of modified frailty index

The accumulating deficits model based on Rockwood’s work of clinical and laboratory factors was used to calculate each patient’s mFI. This frailty index has a long track record in determining the consequences of frailty in medical and surgical studies. A Scopus search using the term “modified frailty index” produced over 1,500 citations, confirming its wide use. Appendix 1 lists the clinical factors and Appendix 2 the laboratory factors used to calculate the mFI. As previously mentioned, not all items listed in Appendix 1 or 2 are necessary to calculate a usable mFI; therefore, only the items that were recorded were used as long as there were at least 5 items. Patient records were reviewed and data points were marked normal, abnormal, or left blank if not recorded. Only items recorded were used in calculating each patient’s mFI. Clinical mFI was calculated as the ratio of the number of deficit items present/total number of items recorded. The laboratory mFI was calculated by the number of abnormal laboratory deficit items present/total number of laboratory items recorded. The combined mFI is the ratio of the combined number of clinical and laboratory deficit items/total number of items recorded. mFI was determined preoperatively based on the admission notes of the physicians and nurses involved in the patient’s care and preoperative laboratory testing. Similarly, mFI for each patient was determined on postoperative day 1 and the day before death or discharge based on the physicians’, nurses’, occupational or physical therapists’ notes, as well as laboratory values. An improvement in frailty was defined as a decrease in mFI score between two time points and was reported as the percentage of patients in each group who showed improvement at that time point.

Statistical analysis

This is a case-control retrospective study design with controls being patients discharged alive and cases being patients with in-hospital postoperative deaths. mFI data is presented as the fraction of the number of items present to the number of items recorded. The change in mFI is the difference between the postoperative day of interest (i.e., postoperative day 1 or final postoperative day) and the preoperative mFI and the difference between the final postoperative day and postoperative day 1 for each patient. A positive number indicates an increase in the mFI, while a negative number indicates a decrease in the mFI. Categorical data are presented as percentages and analyzed using the chi-squared test. Continuous data did not follow a normal distribution and are presented as medians with interquartile ranges. These data were analyzed using the Mann-Whitney U-test. A p-value of 0.05 was considered significant.

Results

100 patients were included in our study, 50 experiencing in-hospital mortality (IHM) and 50 discharged living (DL). Table 1 lists the demographic data of patients suffering a postoperative death and those discharged living. The most common operations among the group were open mitral/aortic valve replacement (n = 45) and partial/total proctectomy/colectomy (n = 27). Among the DL group, the mean age and standard deviation in years was 59 ± 15. The median with interquartile range for length of stay in this group was 7 (4–11) days (Table 1). In the IHM group, the mean age and standard deviation was 64 ± 14 years, with a median with interquartile range length of stay of 10 (4–17) days (Table 1). At each of the three time points, the difference in mFI scores between both groups was statistically significant (Table 2). Patients in the IHM group had significantly greater preoperative mFI scores than those in the DL. This significant difference was present on postoperative day 1, while also increasing in margin (0.240 vs. 0.143, p < 0.00001). The significant difference was present at the final mFI time point while increasing more in margin (0.471 vs. 0.146, p < 0.00001).

Table 1.

Demographics of discharged living (DL) and In-Hospital mortality (IHM)

		Total N	Total Percent %	Discharged Living N	Discharged Living %	In-Hospital Mortality N	In-Hospital Mortality %	DL Mean (SD)	IHM Mean (SD)
Type of Surgery	Open mitral/aortic valve replacement	45	45.0%	15	30.0%	30	60.0%
	Pneumonectomy or pulmonary lobectomy	15	15.0%	12	24.0%	3	6.0%
	Open abdominal aortic aneurysm repair	2	2.0%	1	2.0%	1	2.0%
	Esophagectomy	8	8.0%	5	10.0%	3	6.0%
	Proximal or total pancreatectomy	3	3.0%	3	6.0%	0	0.0%
	Partial or total colectomy or proctectomy	27	27.0%	14	28.0%	13	26.0%
Outcome	Discharged Home	44	44.0%	44	88.0%
	Discharged Not Home	6	6.0%	6	12.0%
	In-Hospital Mortality	50	50.0%			50	100%
Sex	Male	55	55.0%	28	56.0%	27	54.0%
	Female	45	45.0%	22	44.0%	23	46.0%
Age on Admission, mean with standard deviation								59 ± 15	63 ± 14
Length Of Stay, median with interquartile range								7 (4–11)	10 (4–17)

Table 2.

Comparison of preoperative mFI, postoperative day 1 mFI and final mFI of patients surviving and dying after major surgical operations (median with interquartile range)

	Preoperative mFI	POD1 mFI	Final mFI
Discharged Living	0.115 (0.083–0.177)	0.143 (0.089–0.185)	0.146 (0.083–0.215)
In-Hospital Mortality	0.178 (0.121–0.221)	0.240 (0.195–0.302)	0.471 (0.228–0.629)
p-value	0.00009	< 0.00001	< 0.00001

Median Pre-POD1 mFI differences were also significant between the two groups, with operations leading to in-hospital mortality experiencing a greater increase in mFI (0.06 vs. 0.01 p = 0.00019) (Table 3). Median Final-POD1 mFI differences were also significant between the two groups, with IHM patients experiencing a greater increase in mFI while DL patients experienced a slight decrease in mFI (0.148 vs. − 0.005, p = 0.00022). Median Final-Pre mFI differences were also significant between the groups, with a larger margin than the differences between the other two time points (0.276 vs. 0.014, p < 0.00001).

Table 3.

Comparison of the differences between preoperative mFI, postoperative day 1 mFI and final mFI (median with interquartile range)

	POD1-Pre mFI Difference	Final-Pre mFI Difference	Final-POD1 mFI Difference
Discharged Living	0.01 (-0.013–0.061)	0.014 (-0.017–0.046)	-0.005 (-0.022–0.022)
In-Hospital Mortality	0.06 (0.024–0.098)	0.276 (0.079–0.466)	0.148 (-0.020–0.386)
p-value	0.00019	< 0.00001	0.00022

The percentage of patients who had improvement in their mFI scores by each time point was also reported and was higher in the DL than the IHM group at each time point (Table 4). The percentage of patients with Pre to POD1 improvement in mFI was significantly higher in DL than IHM (34% vs. 12%, p = 0.017). The percentage with Pre to Final improvement in mFI was also significantly higher in DL than IHM (32% vs. 16%, p = 0.014). The percentage with POD1 to Final improvement in mFI was higher in DL than IHM and approached statistical significance (52% vs. 28%, p = 0.061).

Table 4.

Comparison of percentage of patients with mFI improvement between discharged living and In-Hospital mortality patients (% of patients with improvement in mFI)

	Pre to POD1	POD1 to Final	Pre to Final
Discharged Living	34%	52%	32%
In-Hospital Mortality	12%	28%	16%
p-value	0.017	0.014	0.061

Discussion

Frailty in older patients has been subjectively acknowledged as carrying increased risk in the perioperative setting [5]. The mFI aims to eliminate the subjective nature of frailty by providing objective scores to quantify patient-associated frailty [10]. This gives the opportunity to further evaluate the impact of frailty in the perioperative period, not only on the older population but also on younger adults. This expands the implications of frailty even further than previously believed, which calls to prioritize frailty as an indicator of patient condition that must be considered when navigating patient care and assessing surgical risk.

The purpose of our study was to explore the concept of dynamic frailty. Namely, that the further accumulation of deficits as enumerated by Rockwood’s accumulating deficits model of frailty will be associated with patients who eventually suffer a postoperative mortality. What is unique to this study is the idea that frailty should not be considered a static characteristic of a patient, but is, in fact, dynamic throughout their hospital course. Frailty, as measured by the mFI, can be assessed periodically during hospital admission to evaluate a patient’s recovery or likelihood of mortality.

Frailty scores based on the mFI continue to accumulate during the postoperative period in patients on a trajectory toward mortality. mFI scores increased (i.e., the patient became more frail) in the IHM group as early as postoperative day 1, and the magnitude in scores increased by the day before discharge or death. In patients who were discharged home or to another facility, the postoperative mFI scores improved in a substantial number of patients, while this was not so for those who died. During the postoperative period, there was a decrease in mFI scores in the patients who were discharged alive, indicating that many patients became less frail in the hospital after their operation. Another way of interpreting this is that the operation improved the patient’s physiological status, which, after all, is what the operation was intended to do. Although more research should be done to evaluate the predictive value of the mFI change from the preoperative state, the statistically significant difference seen in patients’ scores successfully discharged vs. those who expired during their hospital stay is evident with our single institution results.

What we have shown in this study is that the concept of dynamic frailty is worth additional study. Nevertheless, we acknowledge limitations in this study. First we acknowledge that the matching of patients in the DL and IHM groups is not strict. As our purpose was to determine if there was changes within these groups, we do not feel that this detracts from the main aim of the study. Second, there were variations in the number and type of items recorded. Of the 70 possible data points, the mFI requires at least 5 items to be present to provide a reliable index [10]. However, variation in documentation between patient records exists so it is possible that deficits were present and not documented, or they were evaluated in one patient and not another. We also do not know if increasing the number of items used to calculate the mFI would increase its predictive precision. Lastly, we do not know if particular items would contribute more to a patient’s frailty than others. At present, each item is given the same weight. All these are areas of further research.

The potential significant clinical implications of this concept is primarily in the preoperative and immediate postoperative period. Utilizing patients’ pre- and postoperative mFI, evaluation of scores can risk stratify what patients are more clinically unstable and therefore warrant more intensive care. In particular, preoperative scores may better inform providers and patients in operative risk assessment. The mFI’s utility in predicting mortality through preoperative scoring may serve to better identify patients that would benefit from less invasive interventions, therefore avoiding the operating room and the risks involved. While prior studies have examined this association, our data quantifiably displays the deleterious impact surgery has on frailty and its trajectory toward mortality [12, 13]. Data from the mFI can be used to either prioritize modifiable areas of the index to improve clinical stability or put a greater emphasis on palliative care to improve patient comfort given bleak mFI scores. The main practical limitation is documenting the factors used in the mFI calculation, actually calculating the mFI and documenting that calculation in the medical record. In summary, the findings from our study and further research on frailty scores can inform clinical care through improved prognostication and choices of acute intervention or palliative care.

Our study has additional limitations. Firstly, it was a preliminary test of a concept. We purposefully chose operations that have a relatively high risk of mortality. Therefore, how this concept would apply to lower risk operations is an open question. Our prior work demonstrated that even in lower risk operation, the relationship between the mFI and postoperative mortality holds [6]. Secondly, as we randomly selected from a cohort of patients with these operations who died and survived, they were not matched exactly. But as our purpose was to test the concept, we do not feel that this retracts from the results. Thirdly, the mFI was calculated based on the factors that were clearly documented in the medical record as being absent or present. Although as few as five items has been shown to accurately predict adverse outcomes [10], we do not know if the more factors used in the calculation would increase the precision of the predictive value. This is an area of future study. Nevertheless, larger cohorts of more homogenous patients or prospectively designs studies would be needed to confirm these findings.

We believe that this concept of dynamic frailty can be universally applied to many areas of care. It can certainly be applied to non-fatal postoperative adverse events, operation-specific predictions, in young patients, and in non-surgical care such critical care or chronically hospitalized patients. Another area of potential study is the value of each of the factors in determining the mFI. There may be some factors which are not “clinically relevant,” that is, lacking in predictive value. We do not know if such factors exist. These are all areas of potentially fruitful study.

Conclusion

Our data supports previous literature proposing that higher preoperative mFI is a useful predictor of postoperative mortality. This further substantiates evidence for the use of mFI as an indicator of clinical outcomes and an evaluator of patient risk. The mFI consists of a diverse and objective set of data which allows for a holistic evaluation of patients and their vulnerability. The potential for better informed clinical decisions based on frailty evaluations is broad, as is evidenced by the index’s generalizable risk assessment. Preoperative decision making and patient evaluation should incorporate mFI as a consideration in risk-benefit assessments. Moreover, frailty may be used as an effective postoperative tool as early as day 1. This furthers the functionality of mFI as a tool to inform clinical decision making that should not be limited to the preoperative period and highlights mFI’s utility in identifying deteriorating patients who may warrant more intensive care or even palliative considerations. In summary, the evidence for incorporation of mFI as a tool in clinical decision making is strong in both the preoperative and postoperative periods.

Abbreviations

mFI

modified Frailty Index

IHM

In-Hospital Mortality

DL

Discharged Living

PRE

Preoperative

POD1

Postoperative Day 1

Appendix 1

Clinical mFI Items*

System	Deficits
Activities of Daily Living	Problems getting dressed Problems with bathing Problems with personal grooming Problems with cooking Problems with going out alone Difficulty with every day activities
Cognitive/Psychiatric	Short-term memory loss Long-term memory loss Memory changes Difficulty in mental functioning Mood problems Depression Depressed mood Feeling sad, blue or depressed Paranoid features Sleeping disorders Restlessness at night Tired all the time
Neurological	Tremors at rest Action tremors History of Parkinson disease Generalized seizures Impaired vibration Headaches of recent onset Syncope or blackouts History of stroke Clouding or delirium Suck reflex Snout reflex Palmomental reflex Falls
Head and Neck	Head and neck problems Poor muscle tone in neck Facial bradykinesia
Cardiovascular	Arterial hypertension Cardiac symptoms Cerebrovascular problems Decreased peripheral pulses
Respiratory	Respiratory complaints Lung complaints
Gastrointestinal	Gastrointestinal complaints Abdominal complaints Toileting problems
Genitourinary	Urinary incontinence
Endocrine	History of diabetes mellitus History of thyroid disease
Musculoskeletal	Impaired mobility Trunk coordination Poor standing posture Poor muscle tone in limbs Limb bradykinesia Poor limb coordination Gait problems Muscle bulk
Skin and Subcutaneous Tissue	Skin problems Breast problems
Cancer	History of malignancy

^*Deficit if present

Clinical mFI items. The mFI is separated into sets of clinical and laboratory variables meant to assess patient condition. These indices can be further categorized by system (Cardiovascular, Respiratory, Neurological, etc.). For each system, an enumerated set of deficits is evaluated. These measures are then combined with laboratory studies to form the wholistic mFI

Appendix 2

Laboratory mFI items*

Source	Deficit
Physiologic	Blood pressure supine, systolic Blood pressure supine, diastolic
Blood Chemistry	Calcium Creatinine Glucose Phosphorus Potassium Sodium Blood urea nitrogen
Hematologic	Hemoglobin Mean corpuscular volume White blood cell count
Blood Proteins	Albumin Total protein AST (SGOT) Alkaline phosphatase
Vitamins	Folate Vitamin B12
Endocrine	Thyroid stimulating hormone T4
Infectious	Venereal disease research laboratory (VDRL)

^*deficit if present

Data Point	Range
Supine Systolic Blood Pressure	100–140 mmHg
Supine Diastolic Blood Pressure	60–80 mmHg
Calcium	8.5–10.5 mg/dL
Creatinine	0.57–1.11 mg/dL
Phosphorus	2.7–4.5 mg/dL
Sodium	135–148 mEQ/L
Blood Urea Nitrogen	6.0–20 mg/dL
Hemoglobin	12.2–16.2 g/dL
Mean Corpuscular Volume	80–97 fL
White Blood Cell Count	4.6–10.2 × 103/uL
Albumin	3.5-5.0 gm/dL
Total Protein	6.4–8.3 gm/dL
Aspartate Aminotransferase (AST)	5.0–34 u/L
Alanine Aminotransferase (ALT)	5.0–55 u/L
Folate	2.7–17.0 ng/mL
B12	189–833 pg/mL
Thyroid Stimulating Hormone	0.35–4.94 uiu/mL
Thyroxine	0.9–1.7 ng/dL
Venereal Disease Research Lab	negative

The mFI is separated into sets of clinical and laboratory variables meant to assess patient condition. These indices can be further categorized by source (Hematologic, Blood Chemistry, Vitamins, etc). For each source, an enumerated set of values is analyzed for abnormalities. These values are then combined with clinical data to form the wholistic mFI

Author contributions

RM, RR, NA, SK, VV all contributed with study design, data collection, data analysis, manuscript preparation, and have reviewed and approved of the manuscript.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no competing interests.