Incidence and distribution of parkinsonism in Olmsted County, Minnesota, 1976–1990

Study population.

Olmsted County is located in southeastern Minnesota and has a population primarily of northern and central European descent. More than 70% of the population resides in Rochester, the centrally located county seat; the remainder of the county is rural. The local economy is based on farming, health care, and light industry. To be included in the study, patients were required to reside within the county limits at the time of symptom onset. Patients who moved to the county for the management of preexisting parkinsonism were thus excluded.

Case ascertainment.

We ascertained cases through the records-linkage system of the Rochester Epidemiology Project, which provides the infrastructure for indexing and linking essentially all medical information of the county population.8,9 Each provider in the community employs a dossier system (or unit record) whereby all medical information for each individual is accumulated in a single record. Medical diagnoses, surgical interventions, and other key information from medical records are routinely abstracted in a summary record (“master sheet”) and entered into computerized indices using the International Classification of Diseases, Adapted Code for Hospitals (H-ICDA).10

We ascertained potential cases of parkinsonism by searching the indices for the following 53 H-ICDA diagnostic codes: 7 codes for PD, 12 for parkinsonism, 10 for tremor, 8 for other extrapyramidal symptoms, 6 for nonspecific neurodegenerative diseases, 5 for multiple system atrophy, and 5 for progressive supranuclear palsy. A nurse abstractor screened the records of all patients with at least one of these codes entered for the years 1976 through 1995 using a specifically designed form and following a manual of instructions. Although we report incidence findings for the 15-year period from 1976 through 1990, we searched for cases through 1995 to ascertain patients who developed parkinsonism in the study period but were brought to medical attention only later. A neurologist with expertise in movement disorders (J.H.B.) reviewed all records suggestive of questionable or definite parkinsonism and classified patients by types of parkinsonism using a specifically designed clinical data form and following a manual of instructions. The neurologist also determined the year of onset of symptoms.

Diagnostic criteria.

Parkinsonism was defined as the presence of two of four cardinal signs: resting tremor, bradykinesia, rigidity, and impaired postural reflexes. Among patients fulfilling the criteria for parkinsonism, we applied the diagnostic criteria listed in table 1.11-13 Parkinsonism was subdivided into five major types: PD, drug-induced parkinsonism, parkinsonism in dementia, parkinsonism unspecified, and other parkinsonism. Other parkinsonism included vascular parkinsonism, progressive supranuclear palsy, multiple system atrophy, and other rare disorders (e.g., cortical-basal ganglionic degeneration).

Validity and reliability of methods.

We conducted three small-scale validity studies of our case-finding procedures. First, we tested whether our indexing system was sensitive, i.e., whether all diagnoses of parkinsonism listed in the body of the medical dossier were abstracted in the summary record (“master sheet”) and coded properly. As part of a search for controls to be matched by age and sex to our incident cases, we reviewed the complete dossier of 321 individuals who did not have any H-ICDA codes of interest and found no evidence of parkinsonism on or before December 31, 1990 (end of study time window) for any of them. Second, we attempted to estimate whether all patients with parkinsonism in our population came in contact with the medical services of the records-linkage system and were diagnosed. As part of a dementia study in the Rochester population,14 590 individuals underwent a neurologic examination between 1986 and 1997. Of the 16 individuals found to be affected by parkinsonism, 14 were also detected by our case-finding procedure. The remaining two had developed parkinsonism outside our study window (after 1990). Third, we studied the time lag between onset of symptoms and first diagnosis of parkinsonism. Among 360 incident cases with complete information about onset and diagnosis, the median lag time was less than 1 year (median = 0.7; interquartile range = 1.6 years). Our findings suggest that the vast majority of cases of parkinsonism ever detected by the records-linkage system were detected within a relatively short time from onset.

To study the reliability of our case-finding procedure, 20 records of potential cases selected randomly were reviewed independently by two study neurologists (D.M.M. and J.H.B.). For 16 of the 20 individuals (80%) the conclusions of the record review were the same: 5 individuals without parkinsonism, 6 with PD, 2 with drug-induced parkinsonism, and 3 with parkinsonism in dementia. In two cases, the neurologists agreed on the presence of parkinsonism but disagreed on the type (drug-induced versus unspecified in one, unspecified versus PD in the second). Two disagreements were between presence or absence of parkinsonism and were concerning patients whose medical records had only limited documentation of signs and symptoms. For all patients found to be affected by parkinsonism by both neurologists, the agreement on the time of onset of symptoms was within 1 year. Finally, to increase the reliability of our diagnoses, all cases judged problematic by the reviewing neurologist (J.H.B.) were reviewed by a panel including a second movement disorders expert (D.M.M.) and a neuroepidemiologist (W.A.R.). Of 84 cases adjudicated, the panel accepted 74 diagnoses (88%) and revised 10.

Data analysis.

We classified as incident cases all individuals whose record revealed the presence of parkinsonism with symptom onset between January 1, 1976, and December 31, 1990, while they were county residents. We calculated incidence rates using incident cases as the numerator and population counts from the census as the denominator. The denominator was corrected by removing prevalent cases of parkinsonism as imputed from recent European estimates.5 No estimates of the prevalence of parkinsonism in Olmsted County are available since 1965. This correction was done for each calendar year for 5-year age classes between 65 and 99 years and for men and women separately; we assumed stability of these prevalence figures over the 15 years of the study.

We computed age- and sex-specific incidence rates for parkinsonism overall and for specific types. Patterns of distribution were investigated graphically. Age-specific incidence rates were also used to derive cumulative incidence to age 60, 70, 80, and 90 years (assuming no competing causes of death).15

The objectives of this study were mainly descriptive. Because the study covered the target population entirely, no sampling was involved. Statistical tests were, therefore, not appropriate to interpret our findings.16,17

Discussion of methods.

Our study has a number of strengths. By use of the Rochester Epidemiology Project records-linkage system we were able to study retrospectively a large population (1,424,474 person-years overall). Our incidence rates were based on 364 cases of parkinsonism, which allowed for stable rates even after breakdown by age, sex, and type. For most of the incident cases, medical information covered many years of follow-up after the onset of symptoms; many cases (35%) were followed from onset of symptoms through death. This allowed for the use of natural history data (e.g., response to levodopa treatment or appearance of new symptoms or signs) in the differential diagnosis of parkinsonism. Finally, in 265 cases (73%) a diagnosis of parkinsonism was made by a Mayo Clinic neurologist before the time of the study, resulting in extensive documentation of signs and symptoms. However, all patients were rediagnosed using the study criteria to avoid possible changes in diagnostic style over time or across neurologists.

Our study also has a number of weaknesses. The most important potential limitation is the undercounting of parkinsonism cases. Prevalence surveys involving the direct contact of individuals in the general population have repeatedly shown that a sizeable and variable proportion of patients affected by PD are unaware of their diagnosis and have never been diagnosed or treated.3,5,18,19 The implications of these findings for our case-finding strategy are unclear. Although a patient may be unaware of the disease at one point in time, the chance that parkinsonism will be routinely detected at a later time and before death is quite high. Our case-finding strategy covered 20 years of medical records, and we searched the system indices for 5 years after the study period (1991 through 1995). In addition, the Olmsted County population is almost entirely middle-class, well educated, and has excellent access to medical care.9 We conducted three studies to measure the sensitivity of our case-detection strategy; all studies suggested a high level of case ascertainment.

A second potential limitation of our study is the limited reliability of diagnoses obtained through retrospective review of medical records. Our small-scale reliability study showed that the interobserver agreement among two movement disorders specialists was quite good but not complete. The reliability of the diagnosis was increased by using an adjudication panel for all the unusual or controversial cases. Nevertheless, our findings should be interpreted allowing for the possibility of a residual misclassification of cases across specific types.

Discussion of findings.

We found that PD is the most common type of parkinsonism in the overall population. Drug-induced parkinsonism is the second most common type, representing 20% of cases. Parkinsonism in dementia and parkinsonism unspecified are the next most frequent classifications. We show that the distribution by type of parkinsonism changes with age and is different in men and women. Among individuals older than age 79 years, parkinsonism in dementia is the leading diagnosis when considering men and women combined. However, in women, PD remains the most common diagnosis, and parkinsonism in dementia is not as common as it is in men.

We suspect that some of these age and sex differences are due to problems in the application of diagnostic criteria in the context of complex comorbidity. In addition, we designed criteria that were mutually exclusive and did not allow for multiple underlying biological processes. For example, it was difficult to classify demented individuals taking neuroleptics who developed parkinsonism and remained on the neuroleptic through death. Another common problem was classifying patients with apparently concurrent onset of both dementia and parkinsonism. Although we classified patients in whom the cognitive decline predated or was concurrent with the extrapyramidal symptoms as parkinsonism in dementia, many of these cases probably had PD and would present with Lewy body pathology at autopsy.

Neither parkinsonism in dementia nor parkinsonism unspecified represent a unified entity; the pathologic underpinnings of these types are multiple. We have shown that these types represent a significant proportion of patients with parkinsonism, reaching over one-half of all cases among the individuals older than age 79 years. Comorbidities present in this population (e.g., arthritis, peripheral neuropathy, or osteoporosis) often complicate the clinical presentation, precluding the clinician from making a specific diagnosis. Although many cases may actually represent idiopathic PD, others represent different pathologic entities. Further studies of the pathology underlying these diagnostically problematic cases are needed.

The incidence of parkinsonism overall and of the specific types increased steeply with age in both sexes, the only exception being the drop in incidence for the oldest men with PD. This drop is probably an artifact of our diagnostic criteria, reflecting the uncertainty of classification in the presence of comorbidity or of multiple causes of parkinsonism at extreme ages, as discussed above.

The high frequency of drug-induced parkinsonism in our incidence series (20%) is surprising. Prevalence series have estimated a relative frequency of 2 to 9%.3-5 The relative frequency was 7% in the incidence series by Rajput et al.6 Our higher rate of drug-induced parkinsonism may be due to an increased usage of neuroleptics in elderly patients in more recent years.

Higher incidence rates for parkinsonism in men were first reported by Kurland.20 Subsequent studies found no significant differences between men and women.6,7 We found consistently higher incidence rates in men across all age classes and across all types of parkinsonism except drug-induced parkinsonism, which was more common in women. de Rijk et al.5 observed that most prevalence studies indicating sex differences in PD were based on medical records, whereas direct-contact surveys found no sex differences. They suggested that this discrepancy reflects an asymmetric referral of men and women to medical services. However, we have no evidence that such a referral bias is present in our population. A study showed that medical care in Olmsted County is readily available and equal for elderly men and women.9 In addition, the higher rate of drug-induced parkinsonism in women is inconsistent with the lack of referral of women to medical services in this county. Finally, a study of dementia in the same population using a similar case-finding design showed no difference in incidence rates between men and women.21 Therefore, we suggest that the increased rate of parkinsonism in men in Olmsted County is not an artifact. Because parkinsonism is a syndrome with many types, it is likely to have many causes. However, many of the cases of parkinsonism in dementia or parkinsonism unspecified may share the same pathology and presumably the same etiology as PD. This could explain the higher incidence in men across all subtypes of the syndrome. Although still controversial, it has been suggested that hormonal factors could play a role in the etiology of PD.22 However, because of the possible inverse association between cigarette smoking and PD,23 the higher risk of parkinsonism in men may be due to a smoking-related selective mortality. Finally, sex difference in risk may be related to environmental exposures.