History of the Rochester Epidemiology Project
1996; Elsevier BV; Volume: 71; Issue: 3 Linguagem: Inglês
10.4065/71.3.266
ISSN1942-5546
Autores Tópico(s)Genomics and Rare Diseases
ResumoThe Rochester Epidemiology Project is a unique medical records-linkage system that encompasses the care delivered to residents of Rochester and Olmsted County, Minnesota. It is the creation of Dr. Leonard T. Kurland, who envisioned the population-based data resource that would result from combining the clinical documentation developed by the Mayo Clinic with that obtained by other community providers, most notably the Olmsted Medical Group and its affiliated Olmsted Community Hospital. Kurland built on the Mayo unit medical record system that was designed by Dr. Henry S. Plummer in 1907 and on the medical and surgical indexing systems introduced by Dr. Joseph Berkson in 1935. By affording access to details of the medical care given to local residents, the Rochester Epidemiology Project is able to provide accurate incidence data for almost any serious condition and to support population-based analytic studies of disease causes and outcomes. Thus, epidemiologic studies of a wide array of disorders have been possible and have culminated in almost 900 publications since the system was organized in 1966. Olmsted County is one of the few places in the world where the occurrence and natural history of diseases can be accurately described and analyzed in a defined population for a half century or more. The Rochester Epidemiology Project is a unique medical records-linkage system that encompasses the care delivered to residents of Rochester and Olmsted County, Minnesota. It is the creation of Dr. Leonard T. Kurland, who envisioned the population-based data resource that would result from combining the clinical documentation developed by the Mayo Clinic with that obtained by other community providers, most notably the Olmsted Medical Group and its affiliated Olmsted Community Hospital. Kurland built on the Mayo unit medical record system that was designed by Dr. Henry S. Plummer in 1907 and on the medical and surgical indexing systems introduced by Dr. Joseph Berkson in 1935. By affording access to details of the medical care given to local residents, the Rochester Epidemiology Project is able to provide accurate incidence data for almost any serious condition and to support population-based analytic studies of disease causes and outcomes. Thus, epidemiologic studies of a wide array of disorders have been possible and have culminated in almost 900 publications since the system was organized in 1966. Olmsted County is one of the few places in the world where the occurrence and natural history of diseases can be accurately described and analyzed in a defined population for a half century or more. The Rochester Epidemiology Project provides a capability for population-based studies of disease causes and outcomes that is unique in the United States, if not the world. This ability is due to a medical records-linkage system that has afforded access to details of the medical care provided to residents of Rochester and Olmsted County, Minnesota, since the turn of the century. The Rochester Epidemiology Project exists because Olmsted County is isolated from other urban centers, because unit medical records that combine inpatient and outpatient data have been preserved throughout the years, and because indexes to diagnoses, surgical procedures, and test results provide access to the patient records of interest. For exploitation of this situation for epidemiologic studies, however, it was also necessary to relate medical data from the various health-care providers and to develop an infrastructure to support studies using this comprehensive data resource by providing access to expertise in biostatistics, clinical epidemiology, medical informatics, and healthservices research. Herein this review describes the unusual events and extraordinary people who were responsible for the development of the Rochester Epidemiology Project and delineates the unusual research capabilities that have resulted. Subsequent reports will describe the contributions of the Rochester Epidemiology Project to specific clinical problems. The unexpected location of the world's largest private practice of medicine—amid cornfields in the rural Midwest—has its roots in the War Between the States when William W. Mayo moved to Rochester to become examining surgeon for the Enrollment Board, a Civil War induction center. Consequently, the Mayo Clinic ultimately developed in geographic isolation from other urban centers. Today, the closest competing medical centers are in Minneapolis, Minnesota (87 miles to the north), LaCrosse, Wisconsin (71 miles to the east), Iowa City and Des Moines, Iowa (198 and 208 miles to the south, respectively), and Sioux Falls, South Dakota (235 miles to the west). Although best known as a tertiary referral center, the Mayo Clinic has always provided primary and secondary care as well as tertiary care to local residents. Because Mayo offers care in every medical and surgical specialty and subspecialty, local residents are not obliged to seek providers throughout a large region but are able to obtain most of their health care within the community. Indeed, in a 1973 comprehensive survey of community residents, 90% of those who sought medical assessment from a physician in the previous year received such assessment at the Mayo Clinic, the Olmsted Medical Group, or one of their affiliated hospitals, and 96% selected one of these providers when they had a major medical problem.1O'Brien PC. A random survey of Olmsted County, Minnesota, 1973. Technical Report Series, No. 48, Section of Biostatistics, Mayo Clinic, Rochester, Minnesota, 1991Google Scholar More recently, it was found that 83% of the entire Olmsted County population was examined at one or more of these facilities in 1987 (Naessens JM. Personal communication) and that 96% of the hospitalizations among elderly Olmsted County residents in 1988 were at facilities associated with the Mayo Clinic or canvassed by the Rochester Epidemiology Project (Jacobsen SJ. Personal communication). This unusually close correspondence between a circumscribed geographic population and its health-care providers comprises a natural laboratory for population-based studies. These studies are facilitated by the Mayo unified medical record system. In 1907, Dr. Henry S. Plummer introduced the dossier concept of medical records, wherein all data about a specific patient are contained in a single file linked to a unique Mayo Clinic identification number. The ability to incorporate inpatient and outpatient data in these files derives from the exclusive staffing arrangements between the Mayo Clinic and its affiliated hospitals (Saint Marys and Rochester Methodist). These arrangements had their inception when a tornado roared through Rochester in 1883 and leveled much of the town. Difficulties in providing care for the injured persons revealed the need for a hospital. The result was Saint Marys Hospital, built by the Sisters of Saint Francis, nominally a teaching order. Their strong-willed leader, Mother Mary Alfred, initiated construction of the hospital on the condition that Dr. Mayo and his sons (Drs. William J. and Charles H. Mayo) provide medical care. This tradition has continued, and, today, the dossiers for each of the more than 5 million patients who have ever been examined at Mayo Clinic Rochester are maintained in a central repository and tracked by computerized bar code. Because the primary purpose of these unit records is to transfer medical data from one practitioner to another, these records must be readily available, and less than 500 histories have been lost since the unit records were initiated 90 years ago. In order to exploit these clinical data for research, the patient records of interest must be identified. Dr. Plummer recognized this fact and in 1910 devised a diagnostic index (the Plummer-Root file) patterned after the 1888 Nomenclature of Diseases of the Royal College of Physicians. As each patient was dismissed, Plummer's secretary, Mable Root, recorded the date along with the patient's age, gender, and Mayo identification number on a 5-by 8-inch (12.7-by 20.3cm) card that contained up to 216 entries for that particular diagnosis or surgical procedure. The Plummer-Root file supported the first local population study, by Plummer, in 1931.2Plummer HS Exophthalmic goiter in Olmsted County, Minnesota.Trans Assoc Am Physicians. 1931; 46: 171-172Google Scholar His report on the rapidly increasing incidence of exophthalmic goiter in Olmsted County between 1924 and 1927, coincident with the introduction of iodized salt, preceded by several decades the widespread application of epidemiologic methods to the study of noninfectious diseases. As the complexity of medicine increased, however, the strictly anatomic structure of the Plummer-Root file became a handicap. In 1935, Dr. Joseph Berkson introduced a more comprehensive indexing system by using the then-novel Hollerith punch card technology (the Berkson file). This system was based on three-digit codes from the International List of Causes of Death (forerunner of the International Classification of Diseases), which could be collapsed to produce summary reports by body system. Berkson then added sufficient detail so that specific conditions could be identified.3Berkson J A system of codification of medical diagnoses for application to punch cards, with a plan of operation.Am J Public Health. 1936; 26: 606-612Crossref Google Scholar In Berkson's medical index (for diagnoses), this involved adding digits to the base code, anticipating more recent clinical modifications of the International Classification of Diseases. In the surgical index (for procedures), the additional detail was provided by recording the exact procedure that was performed. The Berkson medical index was used until 1975, when it was replaced by a more efficient and flexible on-line computerized system.4Kurland LT Molgaard CA The patient record in epidemiology.Sci Am. 1981 Oct; 245: 54-63Crossref PubMed Scopus (738) Google Scholar Because of increasing volume (approximately 2.5 million diagnoses were indexed in 1995), the medical index is now being further modified for computer-assisted coding of diagnoses in the Section of Medical Information Resources.5Chute CG Yang V Buntrock J An evaluation of computer-assisted clinical classification algorithms.Proc Annu Symp Comput Appl Med Care. 1994; 18: 162-166Google Scholar Berkson's surgical index was used until 1988, when a new surgical information recording system was introduced that is able to record and retrieve the text that describes each procedure. Dr. Leonard T. Kurland, however, was the first to completely exploit this unique potential for generating more accurate frequency and natural history data from population-based studies. Stimulated by a second Olmsted County study, which analyzed the incidence of multiple sclerosis in the community,6MacLean AR Berkson J Woltman HW Schtonneman L Multiple sclerosis in a rural community.in: Woltman HW Merritt HH Wortis SB Hare CC Association for Research in Nervous and Mental Disease. Multiple Sclerosis and the Demyelinaling Diseases. Williams & Wilkins, Baltimore1950: 25-27Google Scholar he came to Mayo for a fellowship in neurology and returned a decade later to become head of what is now the Department of Health Sciences Research. In 1966, he first received funding from the National Institutes of Health to create indexes for the records of the other providers of medical care to Rochester and Olmsted County residents, which would complement those already present at Mayo. The result is linkage of medical data from almost all sources of medical care available to and used by the local population in and near Rochester (Fig. 1), including the Mayo Clinic and its affiliated hospitals, the Olmsted Medical Group and its affiliated Olmsted Community Hospital, the University of Minnesota hospitals and the Veterans Affairs Medical Center, located in Minneapolis, as well as other medical institutions in the region. Only three private general practitioners have had offices in Rochester in recent years, and data from their practices have also been incorporated. These records are available for use in approved research studies. Minnesota law allows release of these data for epidemiologic studies, and protection of patient confidentiality is ensured by stringent Mayo Clinic rules as well as by specific training and guidelines for the handling of such data in the Department of Health Sciences Research. The Rochester Epidemiology Project has been continuously funded for 30 years and was recently renewed through the end of the century. This centralized system now encompasses the medical records of a community population with more than 3.6 million person-years of experience between 1950 and 1995. Olmsted County makes up the Rochester Standard Metropolitan Statistical Area. It includes the central city of Rochester (1990 population, 70,745), which serves as the county seat as well as the center for retail trade and medical care in Olmsted County and the surrounding area. The Mayo Clinic and IBM (International Business Machines) are major employers in the city. Other industry includes food processing and manufacturing of electronic and medical equipment. The rest of Olmsted County (1990 population, 35,725) is rural, with a population density of only 57 per square mile, including the urban fringe of Rochester and one small town. About 1,500 farms in the county are mainly devoted to dairy, com, beef, and swine production. Most of the population is white and of northern European extraction. With the exception of a higher proportion of the working population employed in the health-care industry (24% versus 8% nation ally) and correspondingly higher educational levels, the demographic characteristics of Olmsted County residents resemble those of the US white population (Table 1). Of course, no single community is completely representative of the nation as a whole, but the comparison of previous population-based studies of various chronic diseases in Rochester with those from other communities in the United States indicates that the results for Rochester can be extrapolated to a large part of the US population. For example, for hip fractures, the age-and sex-adjusted incidence among Olmsted County residents 50 years of age or older (385 per 100,000 person-years in 1985 through 1992) was within 3% of the national estimate (394 per 100,000 for the US white population in 1988 and 1989).7Melton III, LJ Atkinson EJ Madhok R Downturn in hip fracture incidence.Public Health Rep. 1996; 111: 17-20Google Scholar The main rationale, however, for studies in this population is not that it is typical but that the existence of the Rochester Epidemiology Project makes such studies far easier to perform.Table 1—Demographic Characteristics in 1990 of Olmsted County, Minnesota, Residents in Comparison With Minnesota and US Residents and US White PopulationOlmsted County (Rochester SMSA)*SMSA = Standard Metropolitan Statistical Area. (%)Minnesota (%)US white population (%)Total US population (%)Race White969410080Gender Female51515151Age (yr) ≥6510.012.513.912.6 Median age31.532.434.432.9Income in 1989 Median per capita$16,214$14,389$15,687$14,420 Below poverty level6.910.29.810.0Education (of ≥25 yr old) High school or more88.082.477.975.2 College or more29.521.821.520.3Industry (of employed persons ≥16 yr old) Agriculture, forestry, fisheries2.64.22.82.7 Mining0.10.30.70.6 Construction4.05.06.56.2 Manufacturing8.418.217.717.7 Transportation2.74.64.34.4 Communication, utilities1.42.02.72.7 Trade (wholesale, retail)27.622.121.221.2 Finance, insurance, real estate4.06.77.26.9 Business, repair services3.04.64.84.8 Personal services3.92.72.83.2 Entertainment0.91.31.41.4 Professional services (health-related services)39.0 (24.5)25.1 (9.4)23.4 (8.0)23.3 (8.4) Public administration2.43.24.64.8* SMSA = Standard Metropolitan Statistical Area. Open table in a new tab During the past 30 years, the Rochester Epidemiology Project, centered in the Section of Clinical Epidemiology, has successfully provided the data and facilities to complete almost 900 reports on the epidemiology of acute and chronic diseases in the population of Rochester and Olmsted County. Several of these investigations have been descriptive studies of disease incidence or prevalence. Much of what is known about the epidemiology of neurologic diseases in particular can be traced to the Rochester Epidemiology Project, inasmuch as many of the initial studies were performed in Rochester. The most unusual feature of these studies is the historical perspective that is possible. With use of the Plummer-Root file for the period 1910 through 1934, the Berkson file from 1935 to 1975, and the computerized system from 1975 onward, long-term secular trends in disease incidence can be assessed. For example, hospital dismissal data from the United States and Canada had suggested that the incidence of hip fracture was increasing among white women (Fig. 2). Age-adjusted data from Rochester during a much longer period indicate that hip fracture incidence rates are no higher now than they were during the early 1940s (Fig. 2). No explanation is known for the rapid increase in hip fracture incidence early during the century, but equally dramatic increases were noted in Scandinavia 2 decades later and more recently in Asia,7Melton III, LJ Atkinson EJ Madhok R Downturn in hip fracture incidence.Public Health Rep. 1996; 111: 17-20Google Scholar Similarly, a study of colorectal cancer among Rochester residents during a 50-year period showed that overall age-and sex-adjusted incidence rates were unchanged between 1940 through 1949 and 1980 through 1989 (47.1 versus 45.7 per 100,000 person-years) and indicated that reports of a doubling in rates during this period could be attributed to underascertainment of colorectal cancer during the early era in other settings.8Beard CM Spencer RJ Weiland LH O'Fallon WM Melton III, LJ Trends in colorectal cancer over a half century in Rochester, Minnesota, 1940 to 1989.Ann Epidemiol. 1995; 5: 210-214Abstract Full Text PDF PubMed Scopus (14) Google Scholar An essential feature of these descriptive studies has been the availability of the original medical records, which allow detailed review to determine whether each potential study subject meets current diagnostic criteria. In the case of malignant lesions and other histologic diagnoses, this includes the ability to reanalzye the original pathologic material inasmuch as formalin-fixed “wet” tissues for Olmsted County residents are available from 1945 to the present, and paraffin blocks and sections are retained indefinitely. For example, slides or blocks were available for review and possible reclassification in 96% of the 613 Rochester residents with lung cancer first diagnosed between 1935 and 1984; thus, we were able to demonstrate that the increase in adenocarcinoma during this 50-year period was not due to changing criteria for histologic diagnosis.9Beard CM Jedd MB Woolner LB Richardson RL Bergstralh EJ Melton III, LJ Fifty-year trend in incidence rates of bronchogenic carcinoma by cell type in Olmsted County, Minnesota.J Natl Cancer Inst. 1988; 80: 1404-1407Crossref PubMed Scopus (28) Google Scholar In addition, the validity of epidemiologic studies depends on complete case ascertainment. Because the original records are available, it is possible to determine whether some patients with the disease of interest were listed under a related diagnosis. An extreme example is shown in Table 2, in which more than 18,000 histories were reviewed to identify 3,105 Rochester residents who fulfilled contemporary criteria for asthma at the Mayo Clinic during a 20-year period.10Yunginger JW Reed CE O'Connell EJ Melton III, LJ O'Fallon WM Silverstein MD A community-based study of the epidemiology of asthma: incidence rates, 1964–1983.Am Rev Respir Dis. 1992; 146: 888-894Crossref PubMed Scopus (636) Google Scholar Although 79% of the patients had been diagnosed with asthma or asthmatic bronchitis, a fifth of all qualifying cases were identified under some other diagnostic rubric. This finding was important because an excess risk of death was noted only among those whose asthma was in conjunction with another lung disorder, primarily chronic obstructive pulmonary disease.11Silverstein MD Reed CE O'Connell EJ Melton III, LJ O'Fallon WM Yunginger JW Long-term survival of a cohort of community residents with asthma.N Kngl J Med. 1994; 331: 1537-1541Crossref PubMed Scopus (125) Google ScholarTable 2—Diagnostic Categories That Produced the Incidence Cohort of Rochester Residents With Asthma*These data relate only to the cases identified from the Mayo portion of the diagnostic index (86% of all incidence cases of asthma).From Yunginger and associates.10Yunginger JW Reed CE O'Connell EJ Melton III, LJ O'Fallon WM Silverstein MD A community-based study of the epidemiology of asthma: incidence rates, 1964–1983.Am Rev Respir Dis. 1992; 146: 888-894Crossref PubMed Scopus (636) Google Scholar By permission.ConditionDiagnosisDefinite asthmaProbable asthmaSingle wheezing episodesTotalCases (%)Asthma or bronchial asthma7762411561,17337.8Asthmatic bronchitis4213615081,29041.5Allergic bronchitis221810501.6Allergic, exogenous, or extrinsic asthma1822220.7Intrinsic, endogenous, or infectious asthma1199290.9COPDf with asthma or bronchospasm642120.4Recurrent bronchiolitis51060.2Wheezing with respiratory infection13941151956.3Wheezing alone1432611073.4Bronchospasm2341811454.7COPD†COPD = chronic obstructive pulmonary disease. in those younger than 30 yr11130.1Chronic bronchitis22150.2Bronchiolitis162215531.7Cough735150.5Total1,4617788663,105100* These data relate only to the cases identified from the Mayo portion of the diagnostic index (86% of all incidence cases of asthma).† COPD = chronic obstructive pulmonary disease. Open table in a new tab In addition to studies of long-term secular trends in disease incidence, comparing the disease experience of the urban residents of Rochester with that of the rural portion of Olmsted County is also possible. For example, in one study, the age-adjusted mortality rate from breast cancer was found to be 40% greater in the urban population.12Melton III, LJ Brian DD Williams RL Urban-rural differential in breast cancer incidence and mortality in Olmsted County, Minnesota, 1935–1974.Int J Epidemiol. 1980; 9: 155-158Crossref PubMed Scopus (13) Google Scholar This increase was partly due to migration after diagnosis, insofar as 20% of those who were rural residents at the time of diagnosis died in Rochester but only 4% of those who were Rochester residents at onset later died in rural Olmsted County; in addition, the incidence of breast cancer was lower in rural Olmsted County, especially in the years before 1955. This system is also notable for including elderly residents, who are underrepresented in medical center series because they are less likely to be referred than are younger persons. For example, the age distribution of patients referred to the Mayo Clinic with Alzheimer's disease differed substantially from that observed among unselected patients from the local community (Fig. 3). The referral patients were a mean of 15 years younger, and 47% of them were men in comparison with only 26% of patients from the community.13Kokmen E Özsarfati Y Beard CM O'Brien PC Rocca WA Impact of referral bias on clinical and epidemiological studies of Alzheimer's disease.J Clin Epidemiol. 1996; 49: 327-335Abstract Full Text PDF PubMed Scopus (90) Google Scholar Other Rochester Epidemiology Project studies have documented comparable distortions in the clinical spectrum that were introduced by referral bias. Thus, reports from academic centers suggested that the typical patient with a pelvic fracture was a young man who had suffered severe trauma and had to be evaluated for potential vascular and urinary tract injuries. Among unselected patients from the community, however, most of those with pelvic fractures were elderly women who had experienced a fall; only 10% of the total had fractures due to severe trauma.14Melton III, LJ Sampson JM Morrey BE Ilstrup DM Epidemiologie features of pelvic fractures.Clin Orthop. 1981; 155: 43-47PubMed Google Scholar The inception cohorts identified in these descriptive studies are frequently used in subsequent population-based historical cohort studies to assess long-term outcomes. For example, the incidence cases of epilepsy identified among Rochester residents between 1935 and 198415Hauser WA Annegers JF Kurland LT Incidence of epilepsy and unprovoked seizures in Rochester, Minnesota: 1935–1984.Epilepsia. 1993; 34: 453-468Crossref PubMed Scopus (1738) Google Scholar provided the basis for retrospective cohort studies of outcomes as diverse as congenital malformation, spontaneous abortion, heart disease, cancer, and fracture.16Annegers JF Hauscr WA Elveback LR Anderson VE Kurland LT Congenital malformations and seizure disorders in the offspring of parents with epilepsy.Int J Epidemiol. 1978; 7: 241-247Crossref PubMed Scopus (97) Google Scholar, 17Annegers JF Buumgartncr KB Hauser WA Kurland LT Epilepsy, anticpileptic drugs, and the risk of spontaneous abortion.Epilepsia. 1988; 29: 451-458Crossref PubMed Scopus (36) Google Scholar, 18Annegers JF Hauser WA Shirts SB Heart disease mortality and morbidity in patients with epilepsy.Epilepsia. 1984; 25: 699-704Crossref PubMed Scopus (130) Google Scholar, 19Shirts SB Annegers JF Hauser WA Kurland LT Cancer incidence in a cohort of patients with seizure disorders.J Natl Cancer Inst. 1986; 77: 83-87PubMed Google Scholar, 20Annegers JF Melton III, LJ Sun CA Hauser WA Risk of age-related fractures in patients with unprovoked seizures.Epilepsia. 1989; 30: 348-355Crossref PubMed Scopus (72) Google Scholar The complete case ascertainment that is possible from review of outpatient and inpatient data often has a pronounced effect on prognosis. Because of complete ascertainment of mild cases, the incidence and prevalence of multiple sclerosis were twice as high in Rochester as previously believed,6MacLean AR Berkson J Woltman HW Schtonneman L Multiple sclerosis in a rural community.in: Woltman HW Merritt HH Wortis SB Hare CC Association for Research in Nervous and Mental Disease. Multiple Sclerosis and the Demyelinaling Diseases. Williams & Wilkins, Baltimore1950: 25-27Google Scholar and the subsequent clinical course was correspondingly more benign than was apparent in referral series. A decade after the onset of multiple sclerosis, most patients were not only alive but also ambulatory and employed. In a recent update that encompassed the 80-year period 1905 through 1984, long-term survival was only modestly reduced from that expected (Fig. 4), and three-quarters of the patients were still alive 25 years after diagnosis. This type of follow-up can be passive, through the linked community medical records, or active. Follow-up activities are consolidated in the Mayo Survey Research Center, which conducted postal, telephone, and face-to-face surveys with almost 40,000 persons in 1995, including both Olmsted County residents and Mayo referral patients. The observed outcomes can be compared with those expected on the basis of the experience of a matched cohort of control subjects monitored in the same way as the cases or from incidence rates for the outcome event that have been established for the general population of Rochester. Thus, among an inception cohort of 1,826 Rochester residents followed up for more than 18,000 person-years from the time of first diagnosis of non-insulin-dependent diabetes mellitus, the risk of a lower extremity amputation was almost 17 times that of Rochester residents in general.21Humphrey LL Palumbo PJ Butters MA Hallett Jr, JW Chu CP O'Fallon WM et al.The contribution of non-insulin-dependent diabetes to lower-extremity amputation in the community.Arch Intern Med. 1994; 154: 885-892Crossref PubMed Google Scholar More than 60% of such amputations in the community were attributable to diabetes. The risk of a transphalangeal amputation was increased almost 400-fold among the diabetic residents, accounting for almost all such amputations in the population. Olmsted County residents with specific disorders may also serve as the cases in case-control studies. These population-based investigations avoid some of the potential biases that often occur in hospital-based case-control studies because the cases represent all incidence cases in the community (rather than a hospitalized subset), because control subjects can usually be selected from among community residents (rather than patients admitted for some other condition), and because exposures can be assessed through long-term medical records. For ascertainment of clinical exposures, this type of study is preferable to interviewing patients, some of whom may have a dementing or critical illness or who may remember past events differently in light of their diagnosis. Thus, putative risk factors were evaluated among 415 Rochester residents newly diagnosed with Alzheimer's disease during 1960 through 1974 in comparison with an equal number of age-and sex-matched community members.22Kokmen E Beard CM Chandra V Offord KP Sehoenberg BS Ballard DJ Clinical risk factors for Alzheimer's disease: a population-based case-control study.Neurology. 1991; 41: 1393-1397Crossref PubMed Google Scholar A review of prior medical records, which had a median duration of 44 years for both cases and controls, revealed significant associations only with a history of hypertension and a few psychiatric disorders. The advantage of comparable ascertainment of risk factors for cases and controls is obvious in a study of patients with dementia but is equally important in all case-control investigations. When important risk factors are identified in population-based studies such as these, estimating their contribution to the overall disease problem in the community is also possible. For example, in another recent case-control study, the combination of diseases associated with secondary osteoporosis or with an increased likelihood of falling accounted for more than 70% of the hip fractures observed among elderly Rochester men.23Poór G Atkinson EJ O'Fallon WM Melton III, LJ Predictors of hip fractures in elderly men.J Bone Miner Res. 1995; 10: 1900-1907Crossref PubMed Scopus (161) Google Scholar This proportion is much higher than was apparent from studies based on interviews with younger patients with hip fractures. Although studies in the Rochester Epidemiology Project have traditionally depended on review of existing medical records, the data system also provides an ability for random sampling of the population for prospective evaluations. Each year, more than half of the Olmsted County population is examined at one of the Mayo facilities, and most local residents have at least one Mayo Clinic contact during any specific 3-year period (Fig. 5). For example, the number of individual women between 65 and 74 years old examined at least once at Mayo during 1989 through 1991 was 96% of the estimated population of Olmsted County in this age-group on the basis of the 1990 census. Additional persons are examined at the Olmsted Medical Group or the Olmsted Community Hospital. Because a very large proportion of the population will have at least one medical contact during some period, the Rochester Epidemiology Project recordslinkage system provides what is essentially an enumeration of the population. Therefore, samples from this system should approximate samples of the general population. This assertion was validated in a recent survey,24Phillips SJ Whisnant JP O'Fallon WM Frye RL Prevalence of cardiovascular disease and diabetes mellitus in residents of Rochester, Minnesota.Mayo Clin Proc. 1990; 65: 344-359Abstract Full Text Full Text PDF PubMed Scopus (229) Google Scholar in which every subject contacted through a random-digit dialing telephone sample or by residence in a local nursing home or senior citizens complex was found to have a medical record in the community. In addition to allowing the selection of community control subjects for medical record studies, such sampling can identify subjects for population-based surveys. One large survey of Olmsted County residents between 30 and 64 years old found that 39% had chronic abdominal pain or disturbed defecation; the prevalence of irritable bowel syndrome alone was 17%.25Talley NJ Zinsmeistcr AR Van Dyke C Melton III, LJ Epidemiology of colonic symptoms and the irritable bowel syndrome.Gastrocnterology. 1991; 101: 927-934PubMed Google Scholar Thus, functional bowel disease was shown to be an extremely common condition, with potentially a substantial effect on the cost of medical care in the community, because 9% of the total Rochester population in this age-group had been examined by a physician for the problem within the previous year. Another area of growing interest relates to the per capita utilization of diagnostic and surgical procedures that can be accessed through the Rochester Epidemiology Project data system. Because new procedures are introduced early at Mayo and rapidly incorporated into practice for the local community, age-and sex-specific utilization rates among Olmsted County residents can be projected to the US population to provide an early estimate of the potential influence of such procedures nationally. For example, we assessed per capita utilization rates for total hip arthroplasty from 1969, when the technology was first introduced in Rochester, through 1990.26Madhok R Lewallen DG Wallrichs SL Ilstrup DM Kurland RL Melton III, LJ Trends in the utilization of primary total hip arthroplasty, 1969 through 1990: a population-based study in Olmsted County, Minnesota.Mayo Clin Proc. 1993; 68: 11-18Abstract Full Text Full Text PDF PubMed Scopus (68) Google Scholar Initially, utilization increased rapidly and then stabilized before increasing again during the late 1980s with the introduction of uncemented implants. Although it is not necessarily true that local utilization is optimal, the previous editor of the New England Journal of Medicine noted that the Olmsted County experience can be viewed as a “conservative but probably good estimate of the medical need” for such procedures.27Relman AS Determining how much medical care we need.N Engl J Med. 1980; 303: 1292-1293Crossref PubMed Scopus (17) Google Scholar The opportunities for such studies are increasing as medical service data are captured in electronic billing files that can be linked for community residents. The potential importance of this capability was demonstrated in a recent study of anemia among Olmsted County residents, which used electronic data from the Mayo Laboratory Information System.28Anía BJ Suman VJ Fairbanks VF Melton III, LJ Prevalence of anemia in medical practice: community versus referral patients.Mayo Clin Proc. 1994; 69: 730-735Abstract Full Text Full Text PDF PubMed Scopus (91) Google Scholar A remarkably high prevalence of anemia was noted among older men and women in the community (Fig. 6). Because 92% of Olmsted County residents 65 years of age and older had a hemoglobin value recorded in this system within a 3-year time frame, we were able to include data from frail and sick elderly persons, who are most likely to have anemia but least likely to participate in surveys, and to document a greater prevalence of anemia in this age-group than found in previous prospective studies. Collecting clinical information in the United States is enormously expensive, and such information typically serves only the immediate needs of patient care. In Olmsted County, however, medical data for the population have been retained and organized into a centralized data system that is available for appropriate clinical, epidemiologic, genetic, and economic studies. The importance of this vast data resource lies in its ability to identify the patient records that provide accurate descriptive information encompassing a half century for almost any disease or syndrome and the ability to conduct population-based analytic studies of cause and outcome. Indeed, the ability to generate long-term incidence, prevalence, and mortality rates has become one of the hallmarks of the Rochester Epidemiology Project. For conditions that do not result in hospitalization for diagnosis, this project may provide the only data available nationally on trends in disease incidence, and thus it functions as an ongoing surveillance system. For example, with the long-term hip fracture rates available in Rochester, we were able to show that no increase in incidence occurred after fluoridation of the Rochester water supply in 1960.29Jacobsen SJ O'Fallon WM Melton III, LJ Hip fracture incidence before and after the fluoridation of the public water supply, Rochester, Minnesota.Am J Public Health. 1993; 83: 743-745Crossref PubMed Scopus (37) Google Scholar In addition, Rochester Epidemiology Project data on the natural history of disease and the long-term outcomes of specific medical interventions have provided urgently needed information on the long-term effects of drugs and other therapeutic modalities. Thus, we were unable to confirm a link between silicone breast implants and subsequent connective tissue disorders.30Gabriel SE O'Fallon WM Kurland LT Beard CM Woods JE Melton III, LJ Risk of connective-tissue diseases and other disorders after breast implantation.N Engl J Med. 1994; 330: 1697-1702Crossref PubMed Scopus (449) Google Scholar Population-based data on health-care needs, services, and costs also are forthcoming. Such information is increasingly needed for evaluating medical care and developing practice guidelines in the Section of Health Services Evaluation. Thus, the Rochester Epidemiology Project is continuing traditions that demonstrate the value of integrating high-quality medical care with education and research, as established by the founders of the Mayo Clinic.
Referência(s)