Yawn BP, Yawn RA, Hodge D, Kurland M, Shaughnessy WJ, Ilstrup D, Jacobsen SJ. A Population-Based Study of School Scoliosis Screening. JAMA. 1999;282(15):1427-1432. doi:10.1001/jama.282.15.1427
Author Affiliations: Department of Research, Olmsted Medical Center (Drs B. Yawn and R. Yawn and Ms Kurland), Sections of Biostatistics (Messrs Hodge and Ilstrup) and Clinical Epidemiology (Dr Jacobsen), and Department of Orthopedics (Dr Shaughnessy), Mayo Clinic, Rochester, Minn.
Context Although school-based screening programs for adolescent idiopathic scoliosis
are mandated in 26 states in the United States, few program outcomes data
exist regarding the effectiveness of such programs.
Objective To determine the effectiveness of a community-based school scoliosis
Design, Setting, and Participants Retrospective cohort study of children who attended kindergarten or
first grade at public or private schools in Rochester, Minn, during 1979-1982.
Children were followed up until age 19 years or until they left the school
Main Outcome Measures Number of children diagnosed and treated for scoliosis, based on results
from scoliosis screenings performed annually in grades 5 through 9, linked
to community medical records data; performance characteristics of the screening
Results Of the 2242 children screened, 92 (4.1%) were referred for further evaluation.
Of these, 68 (74%) had documented medical or chiropractic evaluation of scoliosis.
School screening identified 5 of the 9 children treated for scoliosis but
resulted in referrals for another 87 children who were not treated. The cumulative
incidence of diagnosed scoliosis in this population was 1.8% (95% confidence
interval [CI], 1.2%-2.3%) for curves of more than 10°, 1.0% (95% CI, 0.6%-1.5%)
for curves of at least 20°, and 0.4% (95% CI, 0.1%-0.6%) for curves of
40° or more; 0.4% (0.5% of girls and 0.3% of boys) were treated for scoliosis.
The positive predictive value of the school screening program for the identification
of treated scoliosis was 0.05 (95% CI, 0.048-0.052), with 448 children needed
to screen to identify 1 child who subsequently received treatment. The percent
positive agreement across consecutive years of screening varied from 7% to
Conclusion In this population, school scoliosis screening identified some children
who went on to receive treatment but referred many more who did not. These
data should be considered in making decisions regarding school scoliosis screening.
In the United States, 26 states mandate school scoliosis screening.1 The US Preventive Services Task Force assessed available
data for scoliosis screening but reached no conclusions due to the lack of
randomized controlled trials and observational studies of the outcomes of
US screening programs.1,2 Supporters
of scoliosis screening argue that simple screening tests with adequate sensitivity
and specificity exist and that these programs are necessary to identify scoliosis
early in its natural history, when nonsurgical intervention is possible.3- 9
Opponents emphasize the low incidence of scoliosis requiring treatment, the
resulting low positive predictive value (PPV) of screening programs, and high
numbers of children who would be referred for further evaluation.10- 18
Unfortunately, many of the arguments for and against school scoliosis
screening in the United States are based on mathematical models derived from
small empirical studies. The diagnostic sensitivity and specificity of a single
scoliosis screening test using the scoliometer have been measured.13 However, the effectiveness of scoliosis screening
is best determined by evaluating the complete school screening program, including
the sensitivity and PPV and reproducibility of the series of screening tests
for scoliosis diagnosis and treatment.1,19
Without program outcomes data, it is difficult to assess the value of maintaining
school scoliosis screening programs. To help fill this gap in knowledge, we
evaluated the outcomes of a school scoliosis program in Rochester, Minn.
Rochester is a predominantly white, middle-class metropolitan community
of 70,000 (1990 census) with 1 public and 2 private school systems. Collectively,
the schools serve approximately 1000 students at each grade level.
All children attending public or private schools in Rochester who entered
kindergarten in 1979, kindergarten or first grade in 1980-1981, or first grade
in 1982 and who remained in the Rochester schools for at least 3 of the yearly
school scoliosis screening tests (beginning in grade 5, first year of screening
1984-1985) were included in the cohort. Children were followed up to age 19
years or until they left the school district (latest follow-up, 1994). Children
who were home-schooled (n = 9) were not included.
All Rochester public and private schools participated in the same scoliosis
screening program. Examinations were conducted by public health nurses hired
by the school districts for this purpose. Boys and girls in grades 5 through
9 were screened yearly during physical education class periods. Boys were
asked to disrobe to the waist; girls were allowed to leave their bras in place.
The back was visually inspected for asymmetry. The standard forward bending
test20,21 was performed and beginning
in 1986, a scoliometer was used to quantify rib prominence. Any child with
a scoliometer reading of greater than 6 units or an obvious curve on inspection
was rescreened within 2 to 4 weeks. If the scoliometer reading was again greater
than 6 units or an obvious curve was seen again, a referral letter was sent
to the child's parent(s) or guardian recommending further evaluation by a
health care professional. No parental permission was requested prior to screening.
The screening program was supervised by an orthopedic surgeon who worked with
the nurses in testing and evaluation skills.
Due to parental concern regarding screening examinations that were not
normal but did not meet the criteria for referral, a second type of parental
notification letter, called the "watch letter," was established by the school
district in 1986. This letter notified parents that their children would be
screened every 6 months instead of yearly or reminded parents with children
in the ninth grade that no further school screening would be done. The watch
letter did not specifically recommend professional evaluation.
The results of the scoliosis screening for each child in the cohort
were abstracted from the school health records. The information obtained from
these records included demographic information for each child, the results
of each scoliosis screening in grades 5 through 9, and the dates of any referral
or watch letters sent to the parents. School records are maintained indefinitely
at the central district office for any child who withdraws, transfers, or
graduates; therefore, all children's records were available for review.
All medical records of each child whose parents were sent referral or
watch letters were identified using the record linkage system of the Rochester
Epidemiology Project, which indexes the care provided by all local medical
facilities for all residents of Rochester and Olmsted County.22
To identify all episodes of chiropractic care for these children, records
of all chiropractic clinics in Rochester were searched for any care provided
to any of the referred children and children whose parents received watch
letters. For each child, data were collected for all medical and chiropractic
visits from ages 8 to 19 years.
Data were abstracted from the identified medical and chiropractic records
for each visit in which scoliosis, spinal curve, or school referral was mentioned.
The abstracted data included the date and clinician specialty for all visits;
any evaluations for spinal curvature, including full spine x-rays and the
associated Cobb measurements used to quantify curvatures; recommendations
for further follow-up; and dismissal from care or referrals made, as well
as brace and surgery treatment data. The progression of any documented curve
was recorded, including the greatest degree of curvature before the age of
19 years. Scoliosis treatment was the primary outcome of interest and included
both spinal braces and spinal surgery. For comparability with recent indications
for bracing and surgical treatment of scoliosis,23,24
any documentation of spinal curvatures of at least 20° and 40° or
more was also recorded. For consistency with current orthopedic literature,
a Cobb measurement on a standing full spine x-ray of more than 10° was
used for a diagnosis of scoliosis.4,5,7,8,14,23,24
Children in the cohort who were treated for or diagnosed as having scoliosis
and were not identified through school screening were ascertained by searching
the Rochester Epidemiology Project Diagnostic Index for all diagnoses of "scoliosis"
or "rule out scoliosis." The medical records of these children were also abstracted
in detail. No similar index was available to identify children not referred
from school but evaluated or treated by community chiropractors.
The number of children who did not pass school screening at each grade
level, the number of children who failed multiple times, the percent positive
agreement with future screening results for children who failed at least once,
the number of children who were seen for follow-up care, and the number of
children diagnosed and treated for scoliosis were calculated. The cumulative
incidence of scoliosis in the cohort was estimated by the Kaplan-Meier method,
censoring children who left the community on the last dates of school attendance.
The sensitivity, number needed to screen, and PPV of the school screening
program for identification of children with various levels of scoliosis or
who were treated for scoliosis by 19 years of age were estimated based on
all available data from community medical records.19
Since not all children in the cohort received spine x-rays and the accompanying
Cobb measurement of spinal curve, sensitivity and PPVs were based on available
data. For the primary analysis, a physician's assessment that a child did
not have curvature sufficient to warrant an x-ray was assumed to indicate
no scoliosis. Sensitivity and PPV were calculated in 3 ways to evaluate the
uncertainty of this assumption. These assumed that the referred children who
were not evaluated were (1) all negative for scoliosis, (2) positive for scoliosis
in the same proportion as those who were evaluated, and (3) all positive for
The study protocol was approved by the institutional review boards of
the Olmsted Medical Center, the Mayo Clinic, and the school boards of the
public and private schools in Rochester.
The school records for 2934 children entering school in 1979-1981 were
reviewed. Of this group, 689 (23.5%) left the school district before they
entered seventh grade and were therefore eligible for 2 or fewer school scoliosis
screening tests. Three children were noted to have been diagnosed and treated
for idiopathic scoliosis prior to the first school screening in fifth grade
and were not screened. The remaining group of 2242 children included 92 (4.1%)
children (43 boys [47%], 49 girls [53%]) who were referred 1 or more times
for further evaluation of possible scoliosis. The parents of another 328 children
(14.6%) were sent watch or ninth-grade notification letters (Figure 1).
For many children, the results of screening tests varied from year to
year (Table 1). For example, in
the seventh-grade screening, 9 children failed screening, 8 of whom failed
for the first time. Of the 9 children, 7 were screened in subsequent years
and 2 failed again. The percent positive agreement across consecutive years
of screening varied from 7% to 30% with no pattern of increasing or decreasing
percent positive agreement with the next subsequent test was seen over the
grades 5 to 9.
Medical or chiropractic records could be found for all of the 92 children
referred for additional evaluation. Sixty eight (74%) of the 92 children had
mention of scoliosis evaluation in 1 or more medical records. Of the 68 having
1 or more visits for scoliosis evaluation, 51 (75%) had at least 1 standing
x-ray of the spine. Twenty-three of the 68 were dismissed from scoliosis care
after a single physician visit. The remaining 45 were treated or followed
up. By 19 years of age, 27 (29%) of the 92 referred children (or 40% of the
68 children with scoliosis-related visits) were known to have scoliosis with
curves of more than 10° and 5 (5.4%) were treated (Table 2 [column 1]).
Of the other 328 children whose parents were sent watch or ninth-grade
letters, all but 3 were found to have medical or chiropractic records, and
62 (19%) had notation of a scoliosis assessment. By 19 years of age, none
of these children were treated, and 5 had documented scoliosis of more than
10° (Table 2 [column 2]).
In the school health records, 3 children (1 boy and 2 girls) were noted
as having idiopathic scoliosis identified prior to the first school screening.
Another child was noted to have congenital scoliosis treated before entering
kindergarten and was not included in the analysis. Of the children with idiopathic
scoliosis, 2 were treated with a brace and surgery, and 1 was treated with
a brace only. These children were included in the calculations of community
cumulative incidence and sensitivity of the school screening program.
An additional 9 children in the cohort were found to have had documented
visits for scoliosis evaluation, and 6 were found to have scoliosis. None
were in the groups whose parents received referral or watch letters. The evaluation
of scoliosis was initiated by a physician during a visit for another reason
in 4 children. Visits of the other 5 children were reported to be in response
to school referrals but no parental notification was documented in the school
health records. These 9 children included 1 girl who was treated with a brace
(at 12 years of age) and surgery (at 13 years of age) for scoliosis identified
during a physician visit. She passed scoliosis screening episodes in grades
5, 6, and 7. In addition, 5 other children had a diagnosis of scoliosis, 2
with curves of 11° to 19°, and 3 with curves of 20° to 39°.
By 19 years of age, the Kaplan-Meier cumulative incidence of diagnosed
scoliosis in the entire cohort, determined both from school screening and
community identification, was 1.8% for curves more than 10° (95% confidence
interval [CI], 1.2%-2.3%); 1.0% for curves of at least 20° (95% CI, 0.6%-1.5%);
and 0.4% for curves of 40° or more (95% confidence interval, 0.1%-0.6%).
The cumulative incidence was statistically significantly higher for girls
than boys for curvatures of more than 10° and curves of at least 20°:
2.7% vs 0.9% for more than 10° (P = .003), 1.6%
vs 0.6% for 20° or more (P = .03), and 0.5% vs
0.3% for 40° or more (P = .40).
The sensitivity of the school screening program for identification of
children in the community with treated idiopathic scoliosis was 0.56 (95%
CI, 0.38-0.74) (5/9: 9 children were treated; 3 were identified in the 12
months before the first school screening in grade 5; 1 passed school screening
in grades 5, 6, and 7 and treatment was recommended by age 12 years; and 5
more were treated following referral from school screening). The school screening
program identified 5 (83%) of the 6 children in the cohort who were screened
and subsequently treated for scoliosis. Among the 92 children having a positive
screening examination, 5 were treated. The PPV of the school program was 0.05
(95% CI, 0.048-0.052) (Table 3).
The sensitivity and PPV were higher for the presence of scoliosis of at least
20° or 40° or more. Since not all referred children were evaluated,
a sensitivity analysis was done to estimate the potential range of values
for sensitivity and PPV (Table 3).
The number needed to screen to identify a child who subsequently received
treatment was 448 (2242/5); to identify a child with a curve of at least 20°
by 19 years of age, the number needed to screen was 140 (2242/16). The number
of screening examinations needed to identify each child treated was 2234 based
on the multiple tests children received in grades 5 through 9.
In this cohort of 2242 children, annual scoliosis screening identified
92 children who were referred for further evaluation of possible scoliosis.
An additional 328 parents were notified of possible abnormalities. A total
of 9 children were treated for scoliosis, 5 (56%) of whom were identified
in the school screening program, and 3 were identified prior to the first
During the 5 years of screening, positive screening test results varied
from one year to the next and the low prevalence of treated scoliosis resulted
in a low PPV (5%) for the program and a high number needed to screen to identify
1 child needing treatment (448). The lack of evaluation of 24 of the referred
children may have falsely lowered the PPV; however, it is unreasonable to
assume that any effectiveness study will have 100% compliance with referred
recommendations. If the children who were not evaluated were treated for scoliosis
in the same proportion as those referred and evaluated, the PPV would change
little, increasing to 7%.
Our reported community cumulative incidence of diagnosed scoliosis is
similar to that reported in the literature. In a study of the natural history
of untreated scoliosis, the prevalence of spinal curves was 2.3% for 10°
to 19°, 0.8% for 20° to 39°, and 0.1% for 40° or more at 18
years of age.25 Other reported prevalence rates
range from 0.08% to 4.0% of adolescents, depending on the year of the study,
the definition of scoliosis, the duration of follow-up, the sex and age of
the children studied, and the country in which the study was completed.7,8,11,14,25- 27
We assumed that the 17 children evaluated by physicians for scoliosis and
dismissed from further care without x-rays did not have scoliosis. If some
of these children did have scoliosis, the true cumulative incidence of scoliosis
would be greater than we estimate in this study. In addition, some of the
24 children who had no record of scoliosis evaluation may have had scoliosis
as considered in the sensitivity analysis of the program (Table 3). It is unlikely that all of the 24 children not evaluated
should have been treated for scoliosis. The most clinically reasonable estimate
would be that these children have scoliosis requiring treatment in the same
proportion as the 68 who had a physician evaluation. Under this assumption,
the number of treated children would increase by 2, making the cumulative
incidence of treated scoliosis 0.05 (11/2242). The low incidence of treated
scoliosis led to the discontinuation of school scoliosis screening programs
in several European countries.10,11,14,26
The rate of treatment for children treated in this study is also comparable
to treatment rates (0.2%-0.4%) reported in other studies prior to 1995.14,15 Data from a recent randomized trial
suggest that more children might be considered candidates for treatment today
than during the time period of this study.23
Early brace therapy can slow progression in the 20% to 35% of girls who would
have progression of their moderate (25°-40°) curves.25- 28
It is not possible to know exactly how many children in this study would be
eligible for treatment by current recommendations. The estimated cumulative
incidence of diagnosed scoliosis with curvature of at least 20° suggests
that the upper limit would be approximately 1.0% (95% CI, 0.6%-1.5%). This
is likely to overestimate the number eligible for treatment because some of
the children were skeletally mature when diagnosed or had stable curves of
more than 20° and would not be candidates for bracing. The cumulative
incidence of children who might receive treatment for scoliosis is likely
between the 0.4% treated and the 1.0% with curves of more than 20°.
The marked variability of the screening results deserves special attention.13 Since the scoliometer was not introduced into this
community until the 1986-1987 school year, the criteria for screening failure
changed between the 1985-1986 and 1986-1987 school year. However, all children
were screened in grades 7, 8, and 9 with the scoliometer and the variability
in those years is as marked as the variability seen in grades 5 and 6 (Table 2). Some of the variability could
have been due to the screener's knowledge of prior screening results or knowledge
of ongoing evaluation and treatment occurring outside the school setting.
Treated leg length differences or resolving kyphosis or round back, which
were not noted on the school health record but which allowed the children
to pass future screening tests, may account for some of the variability. Some
children who ultimately proved to have scoliosis by physician and x-ray evaluation
had negative scoliosis screening tests following their first screening failure
and referral. The accuracy of screening assessments in relation to periods
of rapid growth velocity deserves further evaluation, as do our results on
the performance of the series of screening tests.
Not all children who were referred had scoliosis mentioned in their
medical or chiropractic records. The rate of follow-up after receiving a scoliosis
referral letter (74%) is high but lower than the 92% follow-up rate documented
in the same population after referral for failed school vision screening.29 This apparent differential rate of compliance with
school recommendations should be further evaluated since adequate follow-up
is essential for success of a screening program.
In this program, overidentification was compounded by the use of ninth-grade
and watch letters that informed parents of screening results insufficient
to meet referral criteria. These letters resulted in visits for 62 children
who never failed a screening test. None of these children received treatment.
The watch and ninth-grade programs were established at the request of parents.
From the perspective of the school personnel who had no community follow-up
data, a program that notifies 10% of parents regarding a potential health
problem may seem to be a valuable addition to the school screening programs.
These data suggest the watch and ninth-grade letter programs should be discontinued.
Our results may not be generalizable to the entire United States, but
may represent a unique scenario for assessing the value of school scoliosis
screening. No other studies of school scoliosis screening have provided outcomes
data and it is unlikely that a randomized controlled trial of school scoliosis
screening will ever be carried out. The Rochester school districts' student
populations are more than 90% white with income and parental education levels
higher than the state and national norms. The rate of scoliosis is reported
to be higher in the white community than the African American community,26- 28 suggesting that the
screening may appear to be more effective in Rochester than it would in a
racially mixed community. In Rochester, medical care is readily available,
including the practice of 3 pediatric orthopedic surgeons, and more than 85%
of families in the community have health insurance. In addition, the community
was sufficiently concerned about the issue of scoliosis to request the watch
and ninth-grade letter programs. This suggests a high level of awareness and
ability to obtain recommended follow-up evaluation and care. This should provide
an excellent environment to evaluate school scoliosis screening. Conversely,
the same unique community characteristics may have facilitated the detection
of children before or outside of the screening program, making the screening
program appear less useful than it may be in more economically diverse or
underserved communities where disadvantaged adolescents often lack access
to medical care services that are readily available in this community. Students
in disadvantaged communities may lack the opportunity to be identified outside
a school screening program.30
School scoliosis screening identifies some children who ultimately receive
treatment but refers many more who do not. While it is not known how many
children would be identified as having scoliosis without school screening,
a randomized controlled trial of scoliosis screening is unlikely to be completed
in the United States. Therefore, individual communities and state legislators
will need to decide what is appropriate and feasible for their schools based
on the best available data.