Broyles RS, Tyson JE, Heyne ET, Heyne RJ, Hickman JF, Swint M, Adams SS, West LA, Pomeroy N, Hicks PJ, Ahn C. Comprehensive Follow-up Care and Life-Threatening Illnesses Among High-Risk InfantsA Randomized Controlled Trial. JAMA. 2000;284(16):2070–2076. doi:10.1001/jama.284.16.2070
Author Affiliations: Department of Pediatrics, University of Texas Southwestern Medical Center, Dallas (Drs Broyles, Tyson, and Hicks and Ms Hickman); University of Texas-Houston School of Public Health (Drs Swint and Pomeroy); Children's Medical Center of Dallas (Mss Heyne, Adams, and West and Dr Heyne); and Department of Internal Medicine, University of Texas Houston Medical School (Dr Ahn). Dr Tyson is now director of the Center for Population Health and Evidence-Based Medicine, University of Texas-Houston.
Context Inner-city high-risk infants often receive limited and fragmented care,
a problem that may increase serious illness.
Objective To assess whether access to comprehensive care in a follow-up clinic
is cost-effective in reducing life-threatening illnesses among high-risk,
Design Randomized controlled trial.
Setting and Participants A total of 887 very-low-birth-weight infants born in a Texas county
hospital between January 1988 and March 1996 and followed up in a children's
hospital clinic. One hundred four infants who became ineligible or died after
randomization but before nursery discharge were excluded from the analysis.
Interventions Infants were randomly assigned to receive routine follow-up care (well-baby
care and care for chronic illnesses; n = 441) or comprehensive care (which
included the components of routine care plus care for acute illnesses, with
24-hour access to a primary caregiver; n = 446).
Main Outcome Measures Life-threatening illnesses (ie, causing death or hospital admission
for pediatric intensive care) occurring between nursery discharge and age
1 year, assessed by blinded evaluators from inpatient charts and state Medicaid
and vital statistics records; and hospital costs (estimated from department-specific
Results Comprehensive care resulted in a mean of 3.1 more clinic visits and
6.7 more telephone conversations with clinic staff (P<.001
for both). One-year outcomes were unknown for fewer comprehensive-care infants
than routine-care infants (9 vs 28; P = .001). Identified
deaths were similar (11 in comprehensive care vs 13 in routine care; P = .68). The comprehensive-care group had 48% fewer life-threatening
illnesses (33 vs 63; P<.001), 57% fewer intensive
care admissions (23 vs 53; P = .003), and 42% fewer
intensive care days (254 vs 440; P = .003). Comprehensive
care did not increase the mean estimated cost per infant for all care ($6265
with comprehensive care and $9913 with routine care).
Conclusion Comprehensive follow-up care by experienced caregivers can be highly
effective in reducing life-threatening illness without increasing costs among
high-risk inner-city infants.
Neonatal follow-up programs were originally developed to survey the
outcome of high-risk infants, assess the effects of perinatal insults and
care, and identify infants needing referral for care of ongoing problems.
Unfortunately, this approach has often been associated with a substantial
loss to follow-up among families of lower socioeconomic status.1,2
Moreover, this approach does not address the needs of very-low-birth-weight
infants of any socioeconomic situation who lack access to a physician skilled
in managing the pulmonary, gastrointestinal, nutritional, neurological, developmental,
and other problems common among these infants.1- 5
Some follow-up programs now provide well-baby care and care for chronic illnesses.
However, care for acute illnesses typically is not provided. Without prompt,
effective treatment, minor illnesses or complications may quickly become life-threatening
in these vulnerable infants. This problem is likely to contribute to their
increased mortality, morbidity, and cost of care throughout infancy.1,4- 7
Based on our experience in caring for high-risk inner-city infants,1,8- 10 we developed
a comprehensive follow-up program to augment continuity of care and provide
ready access to highly experienced caregivers. Unfortunately, many health
care programs with similar goals have increased costs substantially with little
or no demonstrated improvement in outcome.11- 13
The funding provided to traditional follow-up clinics has been quite limited9,14 and is declining or curtailed in managed
care programs. It is important to demonstrate that the costs of any new follow-up
programs are justified by the benefits, so we conducted a large randomized
trial of comprehensive follow-up care provided by experienced caregivers.
Our purpose was to assess whether providing access to such care is cost-effective
in reducing life-threatening illnesses (illnesses that result in death or
admission to a pediatric intensive care unit) among high-risk inner-city infants.
Our primary hypotheses were that comprehensive care given to high-risk
infants by experienced caregivers would decrease deaths between nursery discharge
and 1-year adjusted age (1 year past term) by 50% and reduce admissions to
a pediatric intensive care unit and total days in such a unit by 33%. (No
reduction in hospitalizations was hypothesized because increased access to
care might increase admissions for all illnesses warranting hospitalization.)
We also hypothesized that comprehensive care would reduce total emergency
department (ED) visits by 33%, reduce the number of infants failing to attend
our clinic at 1 year by 75%, and offset the increased costs for follow-up
care by reducing costs for pediatric intensive care.
Infants born at Parkland Memorial Hospital (the public hospital for
Dallas County, Texas) were eligible if they were born to a Dallas County resident
and either weighed less than 1000 g at birth or weighed between 1001 and 1500
g and received mechanical ventilation in the first 48 hours after birth. Infants
were enrolled within a few days of birth because obtaining parental consent
before the mother's discharge reduces loss to follow-up in the population.1,8,10 Informed consent was
obtained for all participants as approved by our institutional review board.
Sequentially numbered, sealed, opaque envelopes stratified by birth weight
(≤800, 801-1000, and 1001-1500 g) were used to randomize infants to receive
routine or comprehensive follow-up care (if they survived to nursery discharge
and remained eligible for the follow-up clinic). The envelopes were prepared
using a random number table by a person who did not participate in enrollment
The follow-up clinic was located in Children's Medical Center, a private
hospital adjoining Parkland Hospital. Both groups received care from the same
personnel in this clinic: 2 pediatric nurse practitioners and a physician's
assistant, each supervised by both a pediatrician and a neonatologist trained
in follow-up care. These were highly experienced personnel who had a mean
of 11 years' service in our clinic. In each group, infants whose parents spoke
only Spanish were assigned to a bilingual caregiver. In addition to other
responsibilities in the clinic, provider caseloads did not exceed 20 to 30
infants in each group at any time.
Routine Follow-up Care. Routine follow-up care was available 2 mornings per week and included
care for chronic illnesses as well as standard well-baby care (eg, immunizations,
anticipatory guidance, social services, and developmental assessment). All
mothers were taught the signs of acute illness needing prompt evaluation and
told to seek care for these illnesses in neighborhood clinics or the Parkland
Hospital Acute Care Clinic on weekdays or in the Children's Medical Center
ED at other times. (Infants rarely received care in the offices of private
practitioners.) All these sites are staffed by board-eligible or board-certified
pediatricians employed by Parkland Hospital or the University of Texas Southwestern
Medical Center. The neighborhood clinics are part of a well-developed system
of community-based care.15
Measures to maintain contact with the family and encourage attendance
at the follow-up clinic included (1) discussing the clinic with the mother
at study enrollment and shortly before the infant's discharge; (2) scheduling
the first clinic visit within 1 to 2 weeks after nursery discharge; (3) enrolling
the mother and infant in Medicaid before nursery discharge and reenrolling
infants whose coverage had lapsed at any clinic visit; (4) collecting detailed
contact information (including addresses and telephone numbers for the mother
and 2 close friends or relatives) at enrollment that was updated at nursery
discharge and each clinic visit; and (5) contacting the mother to reschedule
any clinic visit that was missed.
Comprehensive Follow-up Care. Comprehensive follow-up care was provided in our clinic 5 days per week
and included care for acute illnesses as well as all components of the routine
follow-up care. The nurse practitioner or physician's assistant who was responsible
for the infant was also available by telephone or pager at all hours to address
acute problems. The primary care clinician contacted 1 of the 2 physician
supervisors as needed and also contacted the ED staff for infants needing
immediate care outside clinic hours. If needed, transportation by taxi was
provided to the ED. On the morning after an ED visit, the mother was called
to assess the need for further evaluation.
To augment rapport, our personnel met with the mother at enrollment
and before the infant's discharge home. A home visit was routinely attempted
except when considered unsafe for our personnel. Mothers thought to have the
greatest need for parenting education and support were offered a trained foster
grandmother of the same ethnic and socioeconomic status.
Follow-up clinic staff did not manage hospitalized infants, determine
who received intensive care, or participate in decisions affecting duration
of intensive care or hospital stay. Study infants were admitted to the hospital
from a variety of sites (the ED, the Parkland Acute Care Clinic, a specialty
clinic, or the follow-up clinic). Whether hospitalization was warranted was
determined by the attending physician. If the infant was seriously ill, this
physician contacted the attending intensivist who was responsible for all
decisions to admit infants to or discharge infants from the pediatric intensive
care unit. Attending physicians were generally not aware of the study, did
not know which infants attending the follow-up clinic were participants, and
were not told the treatment group of study infants.
Admission for pediatric intensive care was used as a marker for the
development of a life-threatening illness. To assess whether an infant died
or received pediatric intensive care before 1-year adjusted age, evaluators
masked to treatment group assessed inpatient medical records for Parkland
Hospital and Children's Medical Center, Texas Medicaid billing records, and
Texas vital statistics records. We were unable to identify whether some infants
had died or survived to 1-year adjusted age because neither a death certificate
nor a record of medical services at or beyond this age was identified. For
these infants, all of the above searches and a search of the outpatient records
were repeated annually in an attempt to find records indicating whether they
had survived beyond 1 year.
Except for 1 hospitalization in another city that was identified through
our clinic records, masked evaluators identified and assessed all inpatient
care using Texas Medicaid and hospital records. Evaluators who could not be
masked to treatment group assessed outpatient care from clinic records at
Children's Medical Center and Parkland Memorial Hospital. Records for the
small proportion of outpatient visits that occurred outside our center were
not reviewed, in part because such visits could not be identified in an accurate
or unbiased manner from Medicaid records or other sources.14
As described previously,14 inpatient
costs were estimated by multiplying hospital charges by department-specific
cost-to-charge ratios from the hospital's annual Medicare Cost Report.16- 18 Outpatient costs
also were estimated using hospital charges and the cost-to-charge ratio for
clinic services. These estimates were adjusted for the increased time that
the comprehensive-care group received from the nursing and physician staff
of our clinic. The salary supplements given primary caregivers for taking
telephone calls after clinic hours also were included in the costs of comprehensive
Our analyses were designed to provide a conservative estimate of the
cost savings from a societal perspective if, as hypothesized, comprehensive
care substantially reduced life-threatening illnesses and admissions for intensive
care. We anticipated that a net cost savings could be verified without evaluating
some costs that would simply reinforce this finding and be difficult and expensive
to assess. For this reason we did not assess indirect costs to the family,
intangible costs (eg, pain and suffering), costs for physician services outside
the follow-up clinic, and costs in the community clinics.
We did assess hospital reimbursements as well as hospital costs. All
estimated costs and reimbursements were inflation-adjusted (using the Consumer
Price Index for Hospital and Related Services), discounted at 3% per annum18 and expressed in 1997 dollars.
The prespecified sample size was 760 infants in the primary analysis
group to allow adequate power to assess the hypothesized effect of comprehensive
care on each primary outcome. The power to assess a 50% reduction in mortality
was 80% (α = .05; expected mortality with routine care, 10%). Enrollment
lasted from January 1, 1988, to March 31, 1996. Hospital costs were assessed
for all infants born after December 31, 1992, when complete costs could first
be obtained for both Children's Medical Center and Parkland Hospital. Time
devoted by clinic personnel to all aspects of care was prospectively assessed
during 6 months (March, April, July, August, and November 1995 and February
1996) considered to be representative of an entire year.
To assess follow-up care between nursery discharge and 1-year adjusted
age, we excluded from the primary analysis group infants who were enrolled
but died before nursery discharge, were discharged after 1 year, or became
ineligible for our clinic because the mother moved out of the county or the
infant was adopted by an out-of-county mother before nursery discharge. These
criteria were considered unlikely to bias the results in favor of our hypotheses,
were selected a priori, and would have been used as exclusion criteria had
enrollment occurred at nursery discharge. All other infants were analyzed
as randomized. None were excluded after nursery discharge.
A 2-tailed Fisher exact test was used to assess categorical variables.
A 2-tailed Wilcoxon rank-sum test was used to assess clinic and ED visits,
admissions, hospital stay, intensive care days, life-threatening illnesses,
costs, and reimbursements. We calculated 95% confidence intervals for the
relative risk of outcomes specified in our hypotheses. Differences between
groups in continuous variables (eg, intensive care days, costs) were not assessed
with confidence intervals because the data were skewed.
During enrollment, 1004 infants met eligibility criteria; 117 could
not be enrolled or randomized (Figure 1).
Before nursery discharge we excluded 51 infants from the comprehensive-care
group and 53 infants from the routine-care group. Thus, the primary analysis
group included 395 infants given comprehensive care and 388 infants given
The groups were at similar high risk (Table 1). More than 80% of the mothers were black or Hispanic, and
their mean education was less than 11 years. The infants had a variety of
major neonatal problems. The comprehensive- and routine-care groups had the
same mean duration of mechanical ventilation (12 days) and virtually the same
mean hospital stay (66 and 65 days, respectively).
On average, the comprehensive-care group had 3.1 more visits to hospital
clinics and 6.7 more telephone contacts with our clinic staff before 1-year
adjusted age than did the routine-care group (Table 2). A foster grandmother was provided for 70 infants given
Infants given comprehensive care were less likely than those given routine
care to cease attending the follow-up clinic by 1-year adjusted age (10.9%
vs 31.4%; P<.001). Moreover, the comprehensive-care
group had fewer total ED visits (597 vs. 730; P =
The comprehensive-care group and routine-care group were similar in
total hospital admissions (273 vs 283; P = .77) and
hospital days (2358 vs 2949; P = .52) (Table 2). All but 5 hospitalizations and 1 admission for pediatric
intensive care occurred at Parkland Hospital or Children's Medical Center.
Despite extensive efforts to contact the families and repeated searches
of medical, Texas Medicaid, and vital statistics records, we could not track
37 infants (4.7%) to 1-year adjusted age and determine whether they survived.
With the increased contact with families, the comprehensive-care group had
fewer such infants than did the routine-care group (9 vs 28; P = .001) (Table 3).
There were 11 deaths identified in the comprehensive-care group and
13 in the routine-care group (P = .68). The groups
were similar in age at and cause of death. Death was unexplained or attributed
to sudden infant death syndrome for 5 infants given comprehensive care and
4 infants given routine care. Death was attributed to infection for 3 infants
given comprehensive care and 4 infants given routine care. Various other causes
were noted for 3 infants given comprehensive care and 5 infants given routine
The comprehensive-care group had 47% fewer infants who developed life-threatening
illnesses (33 vs 62; P = .001) and 56% fewer infants
who received pediatric intensive care (23 vs 52) than did the routine-care
group (P<.001; Table 3). The comprehensive-care group also had 48% fewer total
life-threatening illnesses (33 vs 63; P<.001),
57% fewer admissions to a pediatric intensive care unit (23 vs 53; P = .003), and 42% fewer total days in a pediatric intensive care unit
(254 vs 440; P = .003). In a subgroup analysis excluding
infants who moved from Dallas County after nursery discharge (and were therefore
unable to attend our follow-up clinic), the comprehensive-care group had 70%
fewer days of pediatric intensive care than did routine-care infants (127
vs 418 days; P<.001).
To assess whether the benefits of comprehensive care changed during
the study, we compared the periods before and after May 1, 1992, with respect
to the difference between groups in the proportion of infants with life-threatening
illnesses. In the first period, the routine-care group had 6.7% more infants
with life-threatening illnesses than did the comprehensive-care group ([28/207
infants] − [15/220 infants]). In the second period, the routine-care
group had 8.5% more infants with life-threatening illnesses ([34/181 infants] −
[18/175 infants]). To prevent 1 infant from developing a life-threatening
illness required provision of comprehensive care to 13 infants (calculated
as the inverse of the difference between groups in the proportion with such
The comprehensive- and routine-care groups had similar median hospital
costs for outpatient care ($2165 vs $1944, respectively) and for all care
($2608 vs $2662, respectively). The median value for inpatient costs was $0
for both groups because less than half of the infants in each group were hospitalized
between nursery discharge and 1-year adjusted age.
Mean rather than median costs were used as the appropriate measure of
overall program costs because they indicate the cost impact of reducing infrequent
but highly expensive events. As expected, comprehensive care increased mean
costs for follow-up care (P<.01). However, this
increase was more than offset by the reduction in mean costs for intensive
care beds (P<.03; Table 4). With the hospital accounting methods used, other costs
incurred in the intensive care unit aside from bed costs (eg, laboratory,
radiology, and pulmonary costs) could not be fully separated from corresponding
costs outside the intensive care unit. For all care between nursery discharge
and 1 year, the estimated mean cost per infant was $6265 for comprehensive
care and $9913 for routine care.
Total costs exceeded total reimbursements by an average of $1070 per
infant given comprehensive care and $2997 per infant given routine care. However,
the only statistically significant differences between groups in the economic
analyses were intensive care bed costs and follow-up clinic costs.
In this large randomized trial, access to comprehensive follow-up care
from highly experienced caregivers significantly reduced life-threatening
illnesses and total days of pediatric intensive care by more than 40% among
high-risk inner-city infants. On average, 1 of every 13 infants given comprehensive
care was prevented from developing a life-threatening illness. This finding
indicates a much more favorable "number needed to treat" than for many widely
used medical interventions.19 The benefits
were achieved without increasing total costs. The mean estimated cost per
infant for outpatient and inpatient care between nursery discharge and 1-year
adjusted age was $9913 for routine care and $6265 for comprehensive care.
Adverse outcomes and costs are more likely to have been fully identified
in the comprehensive-care group because these infants had greater contact
with our staff, fewer infants who could not be tracked, and higher maintenance
of Medicaid enrollment14 than in the routine-care
group. Thus, our findings may provide a conservative assessment of the cost-effectiveness
of comprehensive care for the population studied.
No effect on mortality was shown, which may reflect the absence of an
effect, an inadequate sample size to identify the effect, or an inability
to identify all deaths. Despite use of clinic, hospital, and Texas Medicaid
billing records and extensive efforts to contact families, we could not locate
28 infants in the routine-care group and 9 in the comprehensive-care group.
Even in populations at lower risk than in our study, infants who are difficult
to track have a high risk of adverse outcomes.20- 22
The infants in our study were at particularly high medical and social risk
in a population previously shown to have a high prevalence of maternal drug
abuse.23 With no ongoing source of medical
care, some infants who could not be found may have died as a result of untreated
illness, unintentional injuries, neglect, or even homicide.24,25
The better outcome with comprehensive care did not result from inattention
to the routine-care group. The routine-care group (including those who ceased
attending the clinic) had an average of 6.3 clinic visits during infancy and
received care for chronic illnesses as well as standard well-baby care in
the clinic. Care for acute illnesses was available in a well-developed community-based
system of clinics,15 the Parkland Acute Care
Clinic, and Children's Medical Center ED. Thus, access to care for the routine-care
group is likely to have exceeded that for most inner-city infants in the United
States. For such infants, the benefits of a comprehensive-care program might
be greater than we identified. However, it may be difficult to achieve these
results with a less extensive program of comprehensive care or with less experienced
caregivers than in our study.
The difficulty in overcoming adverse socioeconomic factors11,26
was apparent in the finding that 31% of the routine-care group ceased attending
the follow-up clinic. Similar problems occur in other centers serving a similar
population.27 When there has been extraordinary
funding for the follow-up program, low rates of attrition have been achieved
in this population10 and other similar populations
without offering special programs of care. Our findings indicate that providing
comprehensive care can reduce both attrition and adverse outcomes without
increasing total expenditures.
Highly accurate assessments of medical costs require a labor-intensive
and expensive evaluation28 that is impractical,
particularly in multiple hospitals over an extended period. The method that
we used to estimate hospital costs (multiplying hospital charges by department-specific
cost-to-charge ratios) is the most accurate method feasible in most US hospitals.6,16,18 Using this method,
Rogowski6 estimated hospital costs for very-low-birth-weight
infants in California that are comparable to our estimates for the routine-care
group (data available from the authors).
Certain costs were not measured because we anticipated that they would
be expensive and difficult to evaluate and unnecessary to verify the cost-effectiveness
of comprehensive care. These included costs of physician services outside
our clinic, costs for visits to community clinics, and the indirect and intangible
costs (eg, pain and suffering, lost work days) to the infant or family. These
costs are likely to have been systematically higher in the routine-care group—the
group that received all acute care outside our clinic and had substantially
more ED visits and intensive care days—than in the comprehensive-care
group. Thus, our analyses are likely to provide a conservative estimate of
the total cost savings of comprehensive follow-up care from a societal perspective.
Costs are assessed from the hospital's perspective in relation to reimbursements.
In both groups, the estimated costs exceeded Medicaid reimbursements. This
shortfall may discourage hospitals from developing follow-up programs for
high-risk infants. However, the shortfall may be considerably greater in the
absence of any follow-up program. With the effort in our clinic to maintain
Medicaid coverage,14 the estimated shortfall
with comprehensive care ($1070 per infant) was less than with routine care
($2997 per infant).
Because our trial assessed the combined effects of all aspects of the
comprehensive-care program, the effects of the individual components are unclear.
The reduction in life-threatening illnesses was achieved without a substantial
increase in outpatient visits. Indeed, the comprehensive-care group had a
mean of only 3 additional visits to hospital clinics and fewer ED visits than
did the routine-care group. We suspect that the success of the program was
largely due to 24-hour-per-day access to primary caregivers who were highly
experienced in the care of very-low-birth-weight infants. The value of augmenting
continuity of follow-up care and providing prompt attention to acute complications
or illnesses is supported by trials in other high-risk populations.29- 31
The value of comprehensive follow-up care might change with the improving
neonatal outcome of very-low-birth-weight infants. However, there has been
little, if any, change in the proportion of these infants discharged with
chronic lung disease32,33 or neurodevelopmental
deficits,5 problems that increase the likelihood
of life-threatening complications or illnesses during infancy. We found no
decrease over the course of the trial in the benefits of comprehensive care
as measured by the reduction in the proportion of infants who developed life-threatening
illnesses. With the increasing survival of very-low-birth-weight neonates,
the absolute number who would benefit from comprehensive care may be increasing.
For comprehensive care to be effective, follow-up clinics will need
full access to high-risk infants. The best outcomes for high-risk patients
are achieved in high-volume programs designed specifically to meet their needs.34 Our findings support the development of methods to
promote programs of comprehensive follow-up care and the referral of high-risk
infants to these programs. This approach could help improve the outcome of
these infants, reduce the costs of their care, and create better understanding
of the long-term effects of perinatal care and disorders in early life.
In summary, our findings demonstrate that comprehensive follow-up care
provided by highly experienced caregivers can be highly effective in reducing
life-threatening illnesses without increasing the overall costs of care for
high-risk inner-city infants. Follow-up clinics that serve such infants should
consider developing a comprehensive-care program.