The Successful Aging after Elective Surgery (SAGES) Study: Cohort Description and Data Quality Procedures

Abstract

Background/Objectives

Delirium is the most common complication of major elective surgery in older patients. The Successful Aging after Elective Surgery (SAGES) study was designed to examine novel risk factors and long-term outcomes associated with delirium. This report describes the cohort, quality assurance procedures, and results.

Design

Long-term prospective cohort study.

Setting

Three academic medical centers.

Participants

A total of 566 patients age 70 and older without recognized dementia scheduled for elective major surgery.

Measurements

Participants were assessed preoperatively, daily during hospitalization, and at variable monthly intervals for up to 36 months post-discharge. Delirium was assessed in hospital by trained study staff. Study outcomes included cognitive and physical function. Novel risk factors for delirium were assessed including genetic and plasma biomarkers, neuroimaging markers, and cognitive reserve markers. Interrater reliability (kappa and weighted kappa) was assessed for key variables in 119 of the patient interviews.

Results

Participants were an average of 77 years old and 58% were female. The majority of patients (81%) were undergoing orthopedic surgery and 24% developed delirium post-operatively. Over 95% of eligible patients were followed for 18 months. There was >99% capture of key study outcomes (cognitive and functional status) at every study interview and interrater reliability was high (weighted kappas for delirium = 0.92 and for overall cognitive and functional outcomes = 0.94 -1.0). Completion rates for plasma biomarkers (4 timepoints) were 95%-99% and for neuroimaging (one year follow-up) was 86%.

Conclusion

The SAGES study will contribute to the understanding of novel risk factors, pathophysiology and long-term outcomes of delirium. This manuscript describes the cohort and data quality procedures, and will serve as a reference source for future studies based on SAGES.

INTRODUCTION

Delirium, an acute change in attention and cognition, is the leading postoperative complication in older persons, affecting up to half of older patients following surgery. The development of delirium is associated with increased mortality, prolonged functional and cognitive impairment and substantial health care costs, estimated at over $164 billion per year in the U.S.^1-8 Delirium is an important patient safety issue and is increasingly considered as a healthcare quality indicator in older hospitalized patients.^{9, 10} Moreover, delirium is likely to be a major contributor to the ‘post-hospital syndrome’, characterized by prolonged functional impairment following hospitalization and often unrelated to the admitting diagnosis.¹¹ The multifactorial contributors to delirium are often preventable, underscoring the importance of interdisciplinary prospective studies to further elucidate the risk factors, mechanisms, and outcomes of delirium.¹²

The Successful Aging after Elective Surgery (SAGES) study is an innovative, 5-year, prospective program project, funded by the National Institute on Aging, aimed to elucidate novel risk factors (including biomarkers, neuroimaging, and reserve markers) and to examine the contribution of delirium to long-term cognitive and functional decline.¹³ State-of-the-art measurement of delirium and its risk factors and outcomes utilizing quality data collection and management procedures makes the SAGES project uniquely poised to contribute to our understanding of the causes and consequences of delirium.

During the early stages of the study, we published the design and methods¹³ along with several methodological reports validating our key exposure and outcome variables.^14-18 We have now completed enrollment and follow-up to 18 months. The purpose of this manuscript is to provide the baseline characteristics of the SAGES cohort, to describe the procedures used to optimize data collection, and to lay the foundation for future work.

METHODS

Study Population

The SAGES study is an ongoing prospective cohort study of older adults undergoing elective major non-cardiac surgery. The study design and methods have been described previously.¹³ Briefly, eligible participants were age 70 years and older, English speaking, scheduled to undergo elective surgery at one of two Harvard-affiliated academic medical centers and with an anticipated length of stay of at least 3 days. Eligible surgical procedures were: total hip or knee replacement, lumbar, cervical, or sacral laminectomy, lower extremity arterial bypass surgery, open abdominal aortic aneurysm repair, and colectomy. Exclusion criteria included evidence of dementia, delirium, hospitalization within 3 months, terminal condition, legal blindness, severe deafness, history of schizophrenia or psychosis, and history of alcohol abuse. A total of 566 patients were enrolled between June 18, 2010 and August 8, 2013. Written informed consent was obtained from all participants according to procedures approved by the institutional review boards of Beth Israel Deaconess Medical Center and Brigham and Women's Hospital, the two study hospitals, and Hebrew SeniorLife, the study coordinating center, all located in Boston, Massachusetts.

Recruitment and Follow-up

Participant identification, eligibility screening, informed consent, baseline interviews, phlebotomy, and imaging studies were performed on average 2 weeks (mean = 13 ± 15 days) prior to the index surgery. Eligibility was first determined using medical record review, and confirmed by the SAGES team, with formal adjudication by physician investigators if required. Patients were interviewed daily for delirium during the index hospitalization and followed at intervals for at least 18 months and up to 36 months after discharge.

Of 1,052 patients screened for eligibility, 318 (30%) declined to be interviewed. Thus, 734 patients were assessed for eligibility, of which 163 were ineligible and 5 were eligible but refused participation resulting in a total of 566 patients enrolled (Figure 1). The response rate (the percent of the estimated number eligible who were enrolled) was 70% (Appendix A) which is comparable to that observed in many important observational studies.¹⁹ Patients who refused eligibility screening were more likely to be female and less likely to be undergoing orthopedic surgery than patients who agreed to be screened (Appendix Table 1).

Figure 1. Summary of SAGES Enrollment.

Eligible participants were age 70 years and older, English speaking, scheduled to undergo elective surgery with an anticipated length of stay of at least 3 days. Exclusion criteria included evidence of dementia, delirium, hospitalization within 3 months, terminal condition, legal blindness, severe deafness, history of schizophrenia or psychosis, and history of alcohol abuse. Patients who would be out of the catchment area at time of follow-up were also excluded from enrollment.

Data Collection Protocol

Baseline Interviews

A 90-minute baseline interview, conducted before the index surgery in participants’ homes, included detailed data collection on health and functioning. Specific assessments have been described previously.¹³ Key outcomes and other study variables are reviewed below.

Daily In-hospital Delirium Assessment

Delirium was assessed by trained research assessors once daily during the day shift. The Confusion Assessment Method (CAM)²⁰ was rated based on information from patient interviews including a brief cognitive screen (orientation, short-term recall, sustained attention) described previously,¹³ the Delirium Symptom Interview (DSI),²¹ and information related to acute changes in mental status noted by nurses or family members. The CAM is a standardized method for delirium identification with high interrater reliability and sensitivity (94%) and specificity (89%) when compared to clinical expert or consensus-based diagnoses of delirium.²² Delirium severity was assessed using the CAM-S,¹⁵ a validated severity measure based on the CAM that demonstrates strong psychometric properties and strong associations with important clinical outcomes.

Index Hospitalization Medical Record Review

After discharge from the index hospitalization, medical records were reviewed by study clinicians to collect information on surgical procedure, anesthesia type and duration, abnormal laboratory results, baseline diagnoses, development of delirium, precipitating factors for delirium (e.g., medications, iatrogenic events, or catheters), postoperative complications and death. Chart abstraction data were randomly checked for illogical values and against data collected as part of the screening process (e.g., surgery type).

A standardized chart review method for identification of delirium,¹⁴ was used to supplement the CAM interviews. The chart-based delirium diagnosis abstracted information on acute changes in mental status, time and duration of such episodes, evidence of agitation and reversibility or improvement of the acute confusion from in-patient and preoperative notes, discharge summaries, and outpatient visit notes. All chart-based delirium cases were adjudicated by a delirium expert panel and discrepancies were resolved during consensus conferences.

Follow-up Interviews

Home-based interviews were conducted at 1 and 2 months after discharge and then every 6 months up to 36 months. Interviews included assessments of cognitive function, using a neuropsychological test battery (detailed below), and physical function, using the Activities of Daily Living (ADL), Instrumental Activities of Daily Living (IADL) and Short Form 12 (SF-12). Brief phone interviews were conducted with participants at 4, 9, 15, 21 and 27 months after the index hospitalization. Telephone follow-up interviews included a cognitive screen and delirium rating, ADL and IADL assessment, and healthcare utilization since the last interview.

Subsequent Hospitalizations

Information on rehospitalizations was obtained during follow-up interviews with patients and family members. Charts for rehospitalizations were abstracted for delirium and intercurrent illnesses.

Key Study Variables

Delirium

Delirium incidence was defined as presence of delirium according to the CAM criteria during one or more days of hospital interviews or by the chart review method at any time during the hospitalization. Delirium severity was measured using the CAM-S severity score (0-19, 19 most severe) scored from the daily 10-item CAM assessments.

Cognitive and Physical Function

Cognitive function was assessed by a neuropsychological test battery, described previously.¹³ Global cognitive function was assessed using a composite variable created for the study, the General Cognitive Performance (GCP) composite (scaled, 0-100) which synthesized the neuropsychological test results (general population mean = 50, standard deviation =10), and has been demonstrated to be highly sensitive to cognitive change.¹⁶ Physical functioning was assessed by the ADL²³ and IADL,²⁴ and the physical function subscore of the Short Form Health Survey (SF-12).²⁵ A composite physical functioning variable based on these 3 measures was created for the SAGES study, and demonstrated to have strong predictive criterion validity with higher scores on the composite associated with lower risk of discharge to a rehabilitation facility and shorter hospital stays.¹⁷ These outcomes were assessed at baseline, 1 month, 2 months, 6 months and every 6 months thereafter.

Other Study Measures

Race and ethnicity were self-reported by participants. Depression was assessed using the short form of the Geriatric Depression Scale (GDS-15) at baseline, 6 months, and every 6 months thereafter.²⁶ Scores range from 0 to 15, with higher scores reflecting higher severity of depressive symptoms. From the baseline assessment, vision impairment was defined as corrected binocular near vision worse than 20/70 on the Jaeger vision test²⁷ and hearing impairment was defined as hearing six or fewer of 12 numbers out of both ears on the Whisper test.²⁸ The hospital medical record was reviewed to obtain preoperative comorbidity burden measured by the Charlson Comorbidity Index,²⁹ and immediate postoperative severity of illness measured by the Acute Physiology and Chronic Health Evaluation II (APACHE II).³⁰ Participants’ education, occupation, income, participation in cognitively stimulating activities at various points in the life course (18 years, 40 years, etc.) and parents’ education, were collected from participants to operationalize measures of cognitive reserve.

Blood was collected at baseline (prior to surgery), immediately after surgery in the post-anesthesia care unit, on postoperative day 2, and 1-month after surgery. Blood was collected in heparinized (grey top) tubes, placed immediately in ice and transported to the Clinical Research Center, where it was processed within 4 hours of collection. Tubes underwent low speed centrifugation to separate out cellular and plasma components. The cellular component from the baseline time point was used for DNA extraction via the whole blood method.³¹ Plasma was aliquoted into 15 tubes (baseline) or 10 tubes (follow-up) and stored at −80°C to create a SAGES biorepository.

Approximately 25% (n = 146) of enrolled participants elected to undergo brain Magnetic Resonance Imaging (MRI) prior to surgery and 1 year later. Participants who agreed to MRI imaging did not differ significantly from those who did not agree (Appendix Table 2). Exclusion criteria for the nested cohort MRI study included contraindications to 3 Tesla MRI, such as pacemakers, stents, and implants. Participants were imaged on a 3T HDxt MRI (General Electric Healthcare, Waukesha, WI) scanner using a standard eight-channel head coil. The MRI protocol included: Coronal T1, Fluid Attenuated Inversion Recovery (FLAIR), High Resolution 3D T2-weighted imaging, Arterial Spin Labeling (ASL) perfusion, and Diffusion Tensor Imaging (DTI).

Interrater Reliability Assessments

A total of 119 interrater assessments of the in-person interviews were conducted semi-annually. During interrater sessions, two interviewers observed the patient simultaneously but rated the responses independently and were blinded to each other's ratings. All key study variables underwent interrater reliability assessments, including the neuropsychological exam, functional measures (ADL, IADL, SF-12), and delirium. Interrater assessments of medical record abstractions were conducted on a 10% subsample (approximately 60 participants), which involved abstraction of the medical record by two independent raters. For both interview and medical record data, the data stored for future analyses came from one rater that was preselected before any comparison of results.

Data Management and Quality

Data quality procedures were implemented to address four dimensions of quality: completeness (lack of missing data); accuracy (freedom from error); consistency (stability of definitions across databases); and timeliness (sensitivity to temporal change), and were applied to all data collection and management activities.^{32, 33} To optimize accuracy and timeliness, interview staff underwent training including didactics, post-tests, standardization procedures, practice with volunteers, and shadowing of experienced personnel. Completeness and consistency of data capture was assured through database programming and continual cross-checks. Study organizational procedures, including ‘Cores’ responsible for each specific element of data collection (e.g., patient interviews, MRI, phlebotomy)¹³ and team meetings of all Cores resulted in consistent data capture and quality across all measures. The procedures for each data quality dimension across each study activity are described in Table 1. Activities at several stages of data collection and management are described below.

Table 1.

Data Quality Procedures for the Dimensions of Completeness, Accuracy, Consistency and Timeliness

	Completeness	Accuracy	Consistency	Timeliness
Screening and Recruitment
Medical record eligibility screening (EDC)	Flag in database if case not complete (built-in logic)	Computer based eligibility calculations and prevention of most nonsensical data, adjudication	Double check of all eligible cases	Daily tracks of frequency and volumes per site
Recruitment letters and enrollment phone calls	Daily sent letters/phone calls report	Addresses rechecked against United States Postal Service Systems	Standardized letters/phone call scripts	Enrollment date report
In-person and phone interviews, medical record abstraction
Interviewer/rater (data collection level)	Scoring re-check by second person	Team coding/scoring sessions	Interrater reliability interviews	Interview date/ time documentation
Assays
Blood collection	Report of missing samples	Study IDs and dates on all tubes	Time between blood draw and processing time	Standard processing protocol and metrics to ensure timeliness
Neuroimaging
Scan collection	Check for correct number of images per sequence	Visual assessment of coverage and artifacts*	Date alignments with other study dates	Scan date/time documentation
Data Management
Hard copy data entry	Data entry reports	Databases with built-in prevention of nonsensical data entry	Double data entry	Data collection and entry completion reports
Data tracking	Completion and data back-up reports	Checks for out of range or illogical values	Consistency checks of data within participants over time	Out-of-date data collection report
Data cleaning/freezing	Missing data definition and reports	Variable definition sheets#	Independent double-coding of derived variables	Standardized dates to create a ‘frozen’ compendium of validated data (semiannual) used for analyses
Preparation of analysis files	Table shells with mock data for outcomes data	Standardized manual with coding of key variables (‘Code book’)	Data quality report	Report deadlines

^*Artifacts include motion, wrapping/ringing/stripping, blurring, shadowed arc, ghosting, radio frequency noise/spiking, susceptibility artifacts and unexpected inhomogeneities.

^#File with formal definition of the derived variable, description of the handling of missing data, citations of publication first using the variable, and a log of coding decisions.

Participant interviews and medical record abstraction

Reports detailing upcoming interviews were produced weekly. Data collection forms had unique identifiers to prevent misassignment. Submitted forms were checked by independent interviewers to ensure completeness and correct assignment. The data collection team met weekly with study investigators to discuss coding and scoring questions; decisions were recorded in a field operations manual. Interrater reliability assessment was performed semiannually (Table 1).

Blood collection

Phlebotomy tubes were coded with unique identifiers and tracked using the study data management system. Data were regularly cross-checked for alignment with expected dates, and consistent recording of volumes and processing times.

Neuroimaging

Neuroimaging scans were coded with unique identifiers maintained in a separate database and continually cross-checked with the main SAGES database to ensure consistency. Visual assessment of scan quality was conducted in real time and problems were carefully documented.

Data management

Double data entry was performed for paper-based data collection. A codebook with a formal definition of derived variables, description of missing records, chronological records of changes to definitions and coding, and a catalog of ongoing and published analyses relying on each derived metric was regularly updated. A “frozen” compendium of all validated data was created semiannually, and an audit trail of database changes was maintained.

Electronic data were captured with REDCap. Real-time warning messages were triggered when data elements on interviews were left empty. Missing data were queried and reported to staff at weekly meetings, ensuring completeness of the data. Data accuracy was achieved through field validation. Embedded rules, including acceptable values or ranges, prevented erroneous entries.

Statistical Analysis

To describe the cohort, standard descriptive statistics were utilized, including means, standard deviations, proportions and percentages. For interrater reliability assessments, we calculated percent agreement, kappa, and weighted kappa for all interview variables that were assessed (N=119) at item and summary score levels. The kappa statistic is a robust measure of agreement since it assesses agreement occurring beyond chance alone. The weighted kappa statistic is particularly useful when scores are ordered (e.g., more than a dichotomous response) and allows for disagreements to be weighted by degree of disagreement.³⁴

For CAM ratings of delirium, reliability was assessed for the overall rating, and for the 10 individual CAM features. For the overall rating, agreement required exact agreement for the presence or absence of acute change and overall delirium (yes/no). For individual CAM features, agreement was required on the exact level (not present, mild, marked) for each of the 10 features of delirium (e.g., disorganized thinking, altered level of consciousness, etc.). For the neuropsychological test battery, agreement was examined for each individual test. For the Hopkins Verbal Learning Test, agreement for immediate recall required the two raters record the same total number of recalled words over the three trials. For the Visual Search and Attention Task, the two scores were added and agreement on total scores was compared across raters. For Trails A and B, time in seconds was compared and exact agreement was required. For verbal fluency, the total number of words generated for the letters F, A and S were compared and exact agreement was required. For digit span, scores for forward and backwards were added and compared. For the digit symbol substitution test, the category fluency test and the Boston Naming Test, exact agreement on the total correct score was required. For ADLs and IADLs, exact agreement was required for level of dependency (no help needed, help needed, completely unable to perform task) for each task.

RESULTS

Participants were an average of 77 years old (standard deviation = 5.2 years), 58% were female and 93% were white (Table 2). The sample was highly educated, with only 30% having a high school education or less. The majority (81%) of participants underwent orthopedic surgery. At baseline (prior to surgery), 8% of participants reported dependencies in performing ADL, 30% reported dependencies in IADL, and 33% were hearing impaired. The cohort had high cognitive function with an average score on the General Cognitive Composite (GCP) of 58, indicating that the global cognitive function of the cohort, on average, was nearly one standard deviation above the age-matched general U.S. population. Postoperative delirium occurred in 21% of the cohort (n=116) based on the CAM ratings, in 10% (n=57) based on the chart-based ratings and in 24% (n=135) based on the combined CAM and chart-based information.¹⁴ The distribution of missing data at baseline are provided in Table 2.

Table 2.

Baseline Characteristics of Study Cohort (N=566)

Characteristics	n (%)*	Missing Data
Age, mean years (SD)	76.7 (5.2)	0
Female sex	330 (58)	0
Race		0
Asian	5 (1)
African American	29 (5)
White	528 (93)
Other	4 (1)
Hispanic ethnicity	7 (1)	0
Education, mean years (SD)	15.0 (2.9)	0
0-12 Years	165 (29)
13-16 Years	234 (41)
17+ Years	167 (30)
Surgery Type		0
Orthopedic	460 (119)
Vascular	35 (6)
Gastrointestinal	71 (13)
Married	335 (60)
Lives alone	169 (30)	0
Visual impairment, <20/70 corrected binocular vision	3 (.5)	2
Hearing impairment, <6/12 on Whisper Test	185 (33)	1
Geriatric Depression Scale-15, mean score (SD)	2.5 (2.5)	2
Charlson Comorbidity Index, mean score (SD)	1 (13)	0
3MS, mean score (SD)	93.4 (5.4)	1
General cognitive function composite, mean score (SD)	57.5 (7.4)	0
MOS SF-12 physical function composite, mean score (SD)	35.3 (10.0)	3
Any IADL Impairment	157 (28)	0
Any ADL Impairment	42 (7)	0
Postoperative Apache II, mean score (SD)	11.9 (3.0)	0

^*All values are n (%) unless otherwise noted. Proportion of missing values was less than 0.5% on all items.

MOS = Medical Outcomes Study; IADL = Instrumental Activities of Daily Living; ADL = Activities of Daily Living; Whisper: range (0-12), lower is worse; GDS: range (0-15), higher is worse; 3MS:range (0-100), higher is better

Follow-up interviews

Since the study is ongoing, the remainder of this report will focus on the 18-month follow-up, which are completed. In-person or phone interviews occurred 8 times over 18 months following hospital discharge. The dropout rate was less than 1% at each follow-up time point. Participants who dropped out cited time commitment, declining health and memory, or family member concerns with participation as reasons for discontinuation. Between 95% -98% of eligible patients were interviewed at each follow-up point and over 99% of the interviews have complete data on primary outcomes (Table 3). The blood collection rate was between 95% - 99% across the four time points and neuroimaging was completed on 86% of the participants at one-year follow-up (Appendix Table 3). The MRI sequence completion rate was between 90% - 100% at baseline and between 85% - 100% at the one-year follow-up (Appendix Table 4). The variables with the highest percentage of missing data were education level of participants’ father and mother (12% and 9%, respectively), family income (9%) and occupation (6%) (Appendix Table 5). All other study variables were missing in <2% of participants, and the majority were missing in <1%.

Table 3.

Participant Disposition by Study Visit, and Completeness of Data Collection for Selected Major Outcomes

	Participant Disposition					Major Outcomes: Proportion non-missing, %
Visit/Method	Potential Interviews* (N)	Completed Interviews N (%)†	Refused or Unobtainable** (N)	Deaths# (N)	Drop-Outs#	GCP	ADL	IADL
1 Month; In-Person	566	556 (98)	2	1	7	99.6	99.8	99.8
2 Month; In-Person	558	543 (97)	10	1	4	99.4	99.8	99.8
4 Month; Phone	553	537 (97)	10	3	3	--	--	--
6 Month; In-Person	547	537 (98)	10	0	0	99.2	99.6	99.6
9 Month; Phone	547	521 (95)	19	3	4	--	--	--
12 Month; In-Person	540	517 (96)	23	0	0	99.5	100	100
15 Month; Phone	540	502 (93)	32	3	3	--	--	--
18 Month; In-Person	534	508 (95)	26	0	0	99.4	99.7	99.8

Note. Each row describes participants available (eligible) for interview. Reasons for dropout include: time commitment, gatekeeper preference, health status changes, declining memory.

GCP = General Cognitive Composite; ADL = Activities of Daily Living; IADL = Instrumental Activities of Daily Living

^*Eligible defined as all enrolled participants (N=566) minus those who have died or dropped out in the prior time periods

^†Proportion (%) complete defined as the ratio of interviews completed to those potentially available for interview at the relevant visit

^**Those refusing interview may be available at later timepoints

^#Lost to death or follow-up (will not contribute to later timepoints)

Interrater Reliability

Approximately 20 paired interviews were completed per year. Interrater reliability was calculated on individual items and summary scores for the main study variables (Table 4). Percent agreement, kappa and weighted kappa were high for all variables collected. Weighted kappa was 0.92 for overall rating of delirium and ranged from 0.66 (perceptual disturbance) to 0.98 (acute change) for individual features of delirium. For the 4 core features of delirium that are part of the CAM diagnostic algorithm, weighted kappas were all above 0.90, with the exception of disorganized thinking at 0.81 (Table 4).

Table 4.

Interrater Reliability for Key Study Variables

Variable	Agreement %	Kappa	Weighted Kappa
Confusion Assessment Method (N=71)
Acute Change	99%	0.98	0.98
Inattention	93%	0.89	0.92
Disorganized Thinking	92%	0.82	0.81
Altered Level of Consciousness	99%	0.89	0.95
Disorientation	97%	0.92	0.95
Memory Impairment	97%	0.96	0.95
Perceptual Disturbance	97%	0.66	0.66
Psychomotor Agitation	97%	0.74	0.79
Psychomotor Retardation	87%	0.68	0.72
Sleep-Wake Cycle Disturbance	85%	0.77	0.80
Overall Delirium	97%	0.92	0.92
Cognitive and Physical Functioning (N=48)
Neuropsychological Exam
HVLT-R Total Recall	91%	0.91	0.99
HVLT-R Delayed Recall	93%	0.92	0.98
Visual Search and Attention Test	97%	0.97	1.00
Trail Making Test A	95%	0.95	0.97
Trail Making Test B	93%	0.92	0.98
Digit Symbol Substitution	100%	1.00	1.00
Digit Span Test	87%	0.86	0.97
Verbal Fluency	65%	0.64	0.96
Category Fluency	83%	0.81	0.97
Boston Naming Test	100%	1.00	1.00
Overall GCP	98%	0.70	0.94
Basic Activities of Daily Living
Bathing	100%	1.00	1.00
Grooming	100%	1.00	1.00
Dressing	100%	1.00	1.00
Feeding	100%	1.00	1.00
Getting from bed to chair	100%	1.00	1.00
Using the bathroom	100%	1.00	1.00
Walking across small room	100%	1.00	1.00
Dependency Score	100%	1.00	1.00
Instrumental Activities of Daily Living
Telephone	100%	1.00	1.00
Getting places out of walking distance	100%	1.00	1.00
Shopping for groceries	100%	1.00	1.00
Preparing meals	100%	1.00	1.00
Doing housework	98%	0.90	0.90
Managing money	100%	1.00	1.00
Managing Medications	100%	1.00	1.00
Dependency Score	98%	0.93	0.97

Boston Naming range (0-15), higher is better; Dependency Score range (0-14), higher is worse; Digit Span Test range (0-30), higher is better; Digit Symbol Substitution range (0-89), higher is better; GCP = General Cognitive Performance; HVLT-R = Hopkins Verbal Learning Test - Revised; HVLT-R Total recall, range (0-36), higher is better; HVLT-R Delayed recall, range (0-12), higher is better, Trails A range (0-180 seconds), higher is worse; Trails B range (0-300 seconds), higher is worse.

Agreement was calculated for each test in the neuropsychological test battery and the overall GCP (Table 4). Weighted kappas were high, ranging from 0.94 - 1.00. Agreement was 100% for all ADLs and IADLs with the exception of ‘Doing Housework’ with a weighted kappa of 0.90.

DISCUSSION

The study of delirium and long-term cognitive and functional outcomes is inherently challenging. The SAGES study¹³ represents an important advance with a comprehensive preoperative evaluation and longitudinal follow-up of older persons undergoing major surgery. This paper provides the first descriptive summary of the full cohort. The sample includes community-dwelling older persons who had relatively high cognitive functioning at baseline, as measured by neuropsychological testing (e.g., mean of 0.8 standard deviations above the U.S. population mean). The study's data quality procedures, resulted in few losses to follow-up, minimal missing data, and high interrater reliability (kappa =0.92-1.0) on key summary variables. Variables that typically have a high degree of missing data in self-report studies (e.g., income) had <10% missing values.

While there are many longitudinal studies of older adults in the literature, few studies have provided details about their quality assurance procedures. Therefore, this manuscript represents a valuable addition to the literature and serves as a primary reference source for future SAGES work. This paper may provide a useful guide for researchers in the field of aging, and also for clinicians who are trying to identify features of study data quality that might influence their clinical practice.

Strengths of the SAGES study include the measurement of novel risk factors for delirium, including genetic and plasma biomarkers, neuroimaging markers, life-course factors (e.g., early childhood factors such as family income) and reserve markers (e.g., occupational complexity). State-of-the-art approaches were used for measurement of these risk factors, as well as for delirium and long-term cognitive and functional outcomes. In addition to functional outcomes, other important patient-centered outcomes, such as depression and quality of life, were collected longitudinally with little attrition over 18 months. Since SAGES examines a cohort that was free of dementia at enrollment, the study provides a unique opportunity to disentangle the independent contribution of delirium to cognitive and functional outcomes, independent of the effects of pre-existing dementia. The study will provide a complete and valuable data set and biorepository for future work. Finally, the cohort description and data quality procedures will lay the groundwork for future SAGES work.

Several limitations of this study are worthy of mention. Frailty, an important risk factor for postoperative delirium, was not included as part of the original SAGES study; however, we plan to evaluate frailty in our future studies. The sample enrolled patients who were healthy enough to undergo elective surgery, from a single geographic area. In addition, the SAGES cohort was highly educated with limited ethnic and racial diversity. This lack of overall diversity limits the generalizability of findings and requires that the study be replicated with a more diverse cohort. However, these limitations pose no threat to the internal validity of the results. For some outcomes, the study findings might be more conservative in this selected cohort in comparison with a more generally representative sample.

This paper will serve as an important reference source for future studies using the SAGES cohort. We describe the SAGES cohort and data quality procedures to ensure a high retention rate with little missing data for key study variables, and high interrater reliability of all key study variables. The SAGES study will help to advance our understanding of delirium, a complex multifactorial problem and help to clarify risk factors, pathophysiology, and outcomes that ultimately may facilitate the development of targeted prevention and intervention strategies for surgical patients.

Abbreviations

BIDMC

Beth Israel Deaconess Medical Center

BWH

Brigham and Women's Hospital

HMS

Harvard Medical School

HSL

Hebrew SeniorLife

MGH

Massachusetts General Hospital

PI

principal investigator

UCONN

University of Connecticut Health Center