Fork in the road? Why the “mothers-at-home-are-depressed” studies are irrelevant

They pop up with fair regularity in the press, a series of studies purporting to show that mothers at home are unhealthy or depressed. It’s implied that they embody a warning relevant to new mothers, that they offer them a vision of what may befall them depending on the choice they make.

But it turns out that these studies are so riddled with methodological holes as to scuttle any possibility of offering glimpses of alternative futures. The only predictive value they have is to foretell that a woman who stays home by choice risks being averaged in by statisticians with the smaller number of mothers who are at home primarily because they are ill or are engaged in an unsuccessful job search. And that’s just the beginning of the methodological flaws:

(1) The studies don’t compare women at similar stages of life, but rather contrast mothers whose children skew older with those who skew younger

When you read that a study has been done comparing mothers of “young children,” what age cut-off would you imagine was used? Three years old? Six? Ten?

To the Gallup organization, “young children” apparently means anything under eighteen years of age. That was the categorization that gave rise to the recent headline “Stay-at-Home Moms Report More Depression, Sadness, Anger.” The first paragraph reports that the study compared employed and non-employed “women with young children at home.” It takes a bit more reading to discover the generous definition of “young.”

What happens, then, when you take a population composed of women with children under the age of majority and break them into two groups, one at home and one employed for pay? It shouldn’t take a Ph.D. to realize that you get two groups which look significantly different with respect to the age of the youngest child; that the group of at-home mothers is going to skew towards younger children than the working mothers. Far from getting the comparison of staying home versus hiring daycare that the “young children” label would suggest, you’re simply getting confirmation of the common-sense observation that parents of younger children experience more negative emotions because younger children are more challenging to deal with than older ones. (If you like your common sense confirmed by research, a study of 200,000 people in eighty-six countries yielded the conclusion that “New moms and dads who feel overwhelmed by their babies can take comfort in the fact that almost all parents feel happier as they get older.”)

In short, the crystal ball isn’t showing that Future You at Home is likely to be sadder or angrier than Future You at Work at the same stage of life. It’s more like seeing that Future You at home with a two-year-old may be sadder and angrier than Future You at work with elementary-school-aged children. But there’s nothing here to say that Future You leaving a two-year-old at home to go to work would be any less sad or angry.

(2) The studies ding mothers as “depressed” for doing nothing more than reporting the realities of life with an infant – like interrupted sleep

The problem of not controlling for age of the youngest child becomes even more acute when you realize what the studies count as “depressive symptoms.” There is no clinical evaluation here. Mothers are reported as depressed on the basis of questionnaire answers. Oftentimes the at-home and working groups vary by very small margins in their average scores. In any contest in which you win depression points for reporting that your sleep was sometimes restless the previous week, it’s pretty clear that the group with more babies in the house is going to come out ahead.

Which is to say that it is glaringly invalid methodology to count one group as more “depressed” than another on the basis of simple reports of restless sleep when there is every reason to believe that the sleep disturbance doesn’t stem from underlying psychological problems but from external circumstances. This is true regardless of whether restless sleep might be a symptom of depression in some other population. And mothers in these studies not only get depression points if their sleep was disturbed by their infants’ nighttime habits, but bonus points if as a consequence of sleeplessness they check off that they sometimes have trouble “getting going.”

Other questionnaire answers which score as “depressed” similarly reflect the objective realities of life at home with children. “I felt like everything I did was an effort.” “I had trouble keeping my mind on what I was doing.” “I felt lonely.” That accomplishing once-simple tasks is harder with children in tow, that they frequently interrupt what one had one’s mind on – these almost inevitably characterize the part of the day spent in caretaking. The loneliness – well, I’d like to think that part is not inevitable, but the reality is that reduced interaction with adults is an unfortunate feature of being at home in our culture.

Certainly these are trying aspects of being a mother at home. But even a mother who manages these challenges well, who finds the benefits to outweigh the burdens, whose subjective mental health has not tipped over to the negative side according to any reasonable system of classification, is categorized by these studies as depressive simply because she reports honestly that her focus is often interrupted or that more adult companionship would be a boon. The mother with paid employment likely interacts with a wider variety of people who are less disruptive than toddlers. These circumstances may boost her mental health to a higher level – or they might be counterbalanced by other stresses or concerns. But you can’t answer that ultimate question by giving her happiness points simply because she reports that her daily experience includes more people and fewer interruptions.

(3) News flash! Women without health insurance report poorer health!

Suppose you are a woman expecting your first child at age 31. Ideally you would like to have two children spaced a few years apart, and stay home until the younger one is school-aged. You and your husband have saved enough to make this financially possible, and you will continue to be covered under the health insurance your husband gets through his employer.

But you read about a new study on behalf of the American Sociological Association reporting that women who engage in full-time, continuous employment following a first birth have significantly better health at age 40. Should this give you pause?

The answer again is this: Don’t be misled into thinking that such studies are comparing anything like you at age 40 having stayed home for nine years with you at age 40 having worked straight through. Here are the characteristics women had to have in order to be counted in the “stay-at-home” group in the 2012 study: They had to have given birth the first time before age 28, after which they had to have stayed out of the paid workforce for at least 12 years. That means that for twelve years they weren’t earning any money, not even from home (or at least not averaging more than an hour a week at it).

It turns out that when you define stay-at-home moms this way, you end up with a group that is distinguished from working mothers in other important respects. Perhaps the most relevant of these is that about a quarter of the stay-at-home category had no health insurance when they were interviewed by the researchers at age 40. Fewer than 10% of the steadily-working mothers were in that same boat.

So what the researchers were essentially reporting was that a group with a heavier proportion of non-insured women averaged up to have more health problems. When they ran an analysis in which they controlled for the ways the at-home and part-time workers differed from the steadily employed mothers – including the insurance differential – they were “able to reduce” the health differences to “nonsignificance.”

That is to say, this study turned up no health deficit for forty-year old mothers at home so long as they had health insurance. And under the Affordable Care Act, pretty soon all of them will.

Reduced to nonsignificance? That would be a fitting epitaph for this entire genre of research.


 

Here are three of the recent studies specifically referred to in this post, with more detailed analysis for those who are interested:

Elizabeth Mendes, Lydia Saad, and Kyley McGeeney, Gallup, Inc., “Stay-at-Home Moms Report More Depression, Sadness, Anger” May 18, 2012

The researchers controlled for maternal age but not for age of the youngest child in the home. As noted above, this is especially problematic where the age range of children – from birth to age 18 – is so wide. The greater “struggles” of the at-home mothers may reflect nothing more than the fact that their group skewed towards younger children.

Gallup also reports that the stay-at-home mothers were more likely to report having ever been diagnosed with depression. Without further analysis – diagnosed before or after giving birth? – the implication that staying at home causes depression is completely unwarranted. The causation may work in the other direction: women diagnosed with depression may have difficulty maintaining employment and so end up classified as mothers at home.

 

 

Cheryl Buehler and Marion O’Brien, “Mothers’ Part-Time Employment: Associations With Mother and Family Well-Being,” Journal of Family Psychology, 2011, Vol. 25, No. 6, 895-906

This is the study to which Judith Warner recently referred in the New York Times, claiming it showed that “the negative effects of excessive mothering” included “more depression” among mothers at home.

The researchers themselves used the term “depressive symptoms,” a distinction which is often lost in press accounts. (Those “depressive symptoms” are the answers to the “CES-D” questionnaire described in the main blog post, the “Center for Epidemiological Studies Depression Scale.”) Also lost is the researchers’ acknowledgement that “differences between groups were small.”

This study does manage to control in part for age of the youngest child by comparing mothers-at-home and mothers working outside at fixed intervals after the birth of a “focal child” – not necessarily the first child.

But two problems remain. First, the authors make unwarranted assumptions with respect to the direction of causation in the early months, theorizing that their results show employment “protects against depressed mood.” But it may not be the case that mothers at home with six-month-olds or fifteen-month-olds report more disrupted sleep or more difficulty “getting going” than their employed counterparts because the fact of being at home has depressed them. Perhaps the causation is the other way around – that mothers whose babies happen to be fussier, or poorer sleepers, or whose babies may even have significant health problems, tend both to (1) wait longer before returning to work and (2) report on questionnaires that their sleep is disrupted and they have more difficulty with daily tasks.

The second problem is that although the two groups of mothers may tend to be matched for age of the youngest child in the early months, the authors collected no data on the birth of subsequent children. This may explain a particular pattern seen in the results. There were observable differences between the scores of employed mothers after six and fifteen months. Those differences virtually disappeared at 36 months, but were noticeable again at 54 months.

The strong possibility raised by this pattern is that the group still at home at fifty-four months was more likely to have another baby in the house, while for the working mothers the four-and-a-half year old was more likely to be their youngest. Once again, all the implications that the decision to stay home is causing depressive symptoms go completely out the window if the at-home group can score as more depressed just because having younger children leads to more reports of restless sleep.

 

 

Adrianne Frech and Sarah Damaske, “The Relationships between Mothers’ Work Pathways and Mental Health,” Journal of Health and Social Behavior 2012 53:396

This is the study which finds any physical and mental health differences between “steadily” working mothers and stay-at-home mothers reduce to nonsignificance in an analysis which controls for factors including whether there is health insurance.

To its credit, the study distinguishes “interrupted workers” – women who are in and out of the workforce as they experience multiple bouts of unemployment – from other categories, so that they are not lumped in with the at-home grouping or the part-time worker grouping. The interrupted workers do show lower health scores even with controls in place.

This is also the study which defines a mother as being at-home only if she brought in virtually no income for twelve years after the birth of her first child – a qualification met by only about 11% of the women in the study. Women who met this criterion, like part-time workers, averaged higher on a variety of potential stressors at the time they first became mothers, before their mental or physical health could have been in any way impacted by their life with children. Among other distinguishing factors, they were more likely to report prior health concerns limiting their ability to work, less likely to be natives of the United States, and more likely to have given birth as unwed teenagers. When the researchers controlled for these various factors – that is, when they ran statistical analyses comparing at-home mothers and steady workers who had similar metrics at the time of the first child’s birth – the at-home mothers no longer showed any statistically significant mental health disadvantage.

That, despite the fact that they were at that familiar methodological disadvantage of likely having younger children in the house when they were asked to complete their questionnaires at age 40. (It is impossible to confirm this fact as the researchers didn’t ask about or account for the age of the youngest child, but the group of women who had been home twelve years and had significantly more children on average than the steadily-working mothers were also likely to have younger children present.) Once again, questionnaire responses which are presumed to illuminate the impact of employment may be doing nothing more than confirming the mental-health challenges presented by younger children.  This criticism applies not only to the CES-D, with its “restless sleep” questions, which is used in “supplemental models” in this study, but also to the “Short-Form 12” questionnaire on which this study primarily depends. Responders to the “Short-Form 12” are asked how often in the past month they have felt “calm and peaceful” and have “had a lot of energy.” Completely aside from work status, many parents would report that young children tend to drain their energy and detract from calm and peace in their surroundings.

Given their decision to categorize their subjects on the basis of twelve years of work history after the birth of the first child, and their desire to compare them all at the age of forty, the researchers had to exclude women who became mothers after the age of twenty-eight. They reasoned that there was no harm done; that if they had included women giving birth later – “women who delayed motherhood for career and education opportunities” – these women would have been even more likely to stay in the steadily-employed group where their economic and health advantages would have helped that group open up an even bigger advantage over the mothers at home.

I disagree. The question is not whether these women who were older at first birth – women who I agree would on average be educated and likely to engage in significant long-term life-planning – would in many cases have stayed in the workforce and stayed healthy. The question is how big the health differential would have been between the group of later-life mothers who did stay home and the ones who continued to work steadily. It seems to me much more likely that the gap between these two groups would be smaller than the gap between at-home and working mothers who gave birth younger – if for no other reason than that the “planners” were less likely to end up without health insurance. This group would likely have enhanced the health statistics of the mothers at home as well as of the working mothers, in which case excluding them from the study was not as inconsequential as the researchers suggest.


If I had it to do over again, I would still bed-share with my babies

I adopted attachment parenting back in 1986, a few months after our first child was born. A few months before the birth would have been better. That’s why I sometimes envy today’s mothers, who know from the get-go that baby-wearing, sleep sharing, and breastfeeding on cue are viable choices.

On the other hand, parents today have to confront something that I did not: a drumbeat of warnings from the medical establishment that they are endangering their babies’ lives by allowing them in their bed.

Even in the absence of those warnings, it’s not that my husband and I came to bed-sharing easily. Someone had gifted us the 40th anniversary edition of Dr. Spock, and clearly babies spending the night in the parental bed was not something he approved of. He didn’t approve of too much daytime carrying either: “It isn’t necessary or sensible for [the baby] to be in a parent’s lap or arms or to have his father or mother amusing him much of the time.”

So it wasn’t from books we learned attachment parenting; it was from our daughter. She came into the world rosy and strong and pretty clear on what she needed. But we had a crib ready for her and a lot of expectations about babies spending a good deal of the day on their own. (See Dr. Spock, above.) There ensued a struggle. We were certainly not neglectful parents by the standards of the day, but there was a lot of crying going on and—in retrospect—a sort of continual tug-of-war, Elly always wanting to be held and us testing each day to see whether she had turned into that “independent” baby we imagined.

The worst problem, worse even than the long bouts of inconsolable colicky crying, was lack of sleep. It was nearly impossible to get Elly down into her crib without her waking and starting to scream—if not immediately then a very short while later. If we put her in the crib already awake the screaming began the moment she lost contact with us. Dr. Spock said that a half hour of crying would seem like forever, but if we set a timer we would see that it was only 30 minutes after all and that at the end of that time she would almost certainly have fallen asleep.

Not even close.

So we got by on snatches of sleep in the short intervals when we could actually lay her down without waking her. And then in the middle of one night something horrifying happened. Dazed by exhaustion, I failed to get the safety strap latched around her properly as she lay atop her dresser with the built-in changing pad. I took a step away to reach some garment draped over the side of the crib, and turned back just in time to catch her as she rolled off the side. She wasn’t old enough yet to accomplish a purposeful “roll-over,” but she could clearly lurch around with considerable strength.

We were living at the time in a housing project on the Navajo Nation, where my husband and I had gone to practice law. The floor was a hard concrete slab three-and-a-half feet below. I was convinced that Elly had almost died.

After that it no longer seemed worth braving sleep deprivation in order to uphold some notion of where babies should sleep, and we brought her into our bed. We worried at first about rolling over on her. But I realized that she was not too helpless to raise an alarm if someone started leaning on her, and it really was not too likely that we were going to land on her suddenly as is from a great height—any more than my husband and I were likely to fall out of bed or slug one another in our sleep.

Looking for reassurance, I would eventually find Dr. Sears’ early writings, and when we left the Reservation some months later I found La Leche League enormously supportive. Meanwhile I took some comfort in the example of the co-sleeping Navajo mothers around me, and even the Pakistani co-workers who had pushed the mattresses together in their bedroom to create a wall-to-wall bed for the family.

Our new arrangement allowed me enough sleep to function, definitely more than we had gotten before, but I still wouldn’t say we slept all that well. When baby Matthew arrived three years after his sister, he was in our bed from day one, and the adjustment went much more smoothly. This is not to say we slept through the night unbroken, but we achieved a synchronicity where we would both emerge from sleep together to nurse, and then drift off again.

Now, though, that a lot of parents have gotten past the old bugaboos about the psychological risks of bed-sharing, we have a host of new warnings that bed-sharing is dangerous, from organizations including the American Academy of Pediatrics. The experts would have us believe that my husband and I, waking every morning to find our children alive and well, had dodged a bullet. Or perhaps it was that we had dodged a knife – that being the implement shown lying next to a baby in a public service announcement run by health authorities in Milwaukee. The message was that if you wouldn’t allow your infant to sleep next to a cleaver, you shouldn’t allow him or her to sleep next to an adult either.

I am the kind of conscientious person who takes cautionary statistics and public health campaigns very seriously. But at some point in my life I realized that the fact [oops! turns out it’s a myth] that half of all marriages ended in divorce didn’t mean that mine faced fifty-fifty odds; I had inside information allowing me to make a more customized assessment of the situation. And when the state of Arizona phased in a requirement for hepatitis B vaccinations among middle-schoolers in 2000, it made no sense for us as a family to vaccinate our pre-teen son but not our teen-aged daughter, for all that that was the plan the bureaucrats were promoting.

It is precisely because the authorities (unfortunately) see it as their job to come up with one-size-fits-all prescriptions that making an informed decision to ignore their advice may be the most responsible thing a parent can do. For a long time the warnings against bed-sharing issued by the American Academy of Pediatrics were based on studies which folded into the statistics parents who smoked, parents who came to bed inebriated or otherwise incapacitated, and parents who spent the night with their babies on a couch. Parents who knew—as did my husband and I—that there was no chance they were going to fall into bed dead drunk a few weeks after initiating the bed-sharing habit would be justified in concluding that the research results had very little to do with them, regardless of what behavior the medical authorities might attribute to their audience.

In mid-2013, though, a journal article was published which re-analyzed older data and reached the conclusion that for infants under three months old, bed-sharing even under relatively safe conditions is correlated with an elevated risk of Sudden Infant Death Syndrome (SIDS). Elevated, that is, by comparison to what the AAP considers the gold standard of infant sleep—breastfed babies sleeping on their backs in cribs in their parents’ bedroom. (Formula-fed babies and babies sleeping in rooms apart from their parents show higher rates of SIDS—for all that there are no public service announcements depicting the Grim Reaper stalking the nursery.)

So had the research finally caught up to the recommendations, retroactively justifying the medical establishment’s warnings?

Not quite.

The methodology of these studies is extremely tricky to design, and there was at least one major pitfall which the 2013 analysis had failed to overcome. It didn’t distinguish between babies for whom the parental bed was their primary sleep location, and babies who were only occasionally in the parental bed.

Let me hasten to explain that the concern here has nothing to do with whether families who co-sleep only part of the time are somehow more awkward at it when they do. The problem is a statistical one which confounds the researchers’ attempts to sort out cause and effect.

In carrying out these “case-control” studies, the researchers look to see what percent of babies who died of SIDS were in the parental bed at the time, and then compare that to the rate of co-sleeping among similar “control” babies. If the proportion of co-sleepers was higher among the babies who succumbed, then the researchers conclude that the co-sleeping itself created some element of danger.

But what if the reason more of the SIDS babies were co-sleeping is that their parents sensed something wrong with them? That is, how do we know that there wasn’t some factor pertaining to the health of the baby which both increased the risk of SIDS and increased the probability that the parents would take the infant into the bed?

Something, after all, determines whether the infant in a flexible-sleeping-arrangement household is in the parental bed at any given moment. It would take only a small number of cases in which that something was an infant facing a health challenge to undermine the researchers’ conclusions. It wouldn’t even be necessary that the parents were consciously aware of the baby being ill; just that something about the baby’s behavior—fussiness, or difficulty settling—upped the likelihood that he or she would be taken into the parental bed.

The only way that I can see to avoid this methodological flaw would be to break out primary co-sleeping families, those in which the baby routinely spends the better part of every night in the parental bed.

There are indeed studies which talk about having analyzed “usual” or “routine” co-sleepers. But on further examination it turns out these researchers still weren’t looking at primary co-sleepers, but using far vaguer, even inconsistent, definitions of “routine.” Take for example a 2009 German study using one of the same databases—and with one of the same authors—as the 2013 meta-analysis. Of 333 AIDS cases, only 26 families responded “parental bed” when asked where their babies usually slept. Yet one table later, there are 146 SIDS cases classified as “usual” co-sleepers. It turns out this referred to infants who were reported to spend some interval co-sleeping either “sometimes” or “every night,” as opposed to “never.” In other words, the researchers had identified exactly the kind of flexibly-sleeping family whose infant was in the parental bed when there was some reason for him or her to be there. The fact that 27% of the infants who succumbed to SIDS in these “sometimes” bed-sharing families were in the parental bad at the time of death, while only 22% of the “sometimes” bed-sharing controls were in the parental bed during the last sleep inquired about, gave rise to the conclusion that even “usual” bed-sharers were endangered by being the parental bed.

What, though, about the 26 families who reported the parental bed as their infants’ primary sleep location? That group showed no increased risk of SIDS.

It remains a distinct possibility that for infants whose parents don’t drink, drug, or smoke, the family bed is the safest sleeping arrangement of them all.

The 27% vs. 22% comparison should give an idea that the differences the researchers are seeing between the groups they contrast are not large. In the 2013 meta-analysis, the authors set out to quantify for readers how much additional risk was being run by bed-sharing families. They indicated that in a typical situation, room-sharing babies laid on their backs in their cribs by non-smoking parents would be expected to succumb to SIDS in about 8 out of 100,000 cases, while for bed-sharing cases the number would rise to about 23 in 100,000.

But there turned out to be something a bit odd about that “typical.” The numbers provided assumed a baby born weighing between 2.5 kg (5 pounds 8 ounces) and 3.5 kg (7 pounds 11 ounces). This included the low end of normal birthweights, but lopped off the entire upper end.

At least one-third of babies weigh more than 7 pounds 11 ounces at birth and are at markedly decreased risk of SIDS no matter where they sleep. In fact among the control babies in the study who were supposed to represent the general population, fully half weighed over 3.5 kg at birth, calling into question whether the numbers the researchers chose to highlight really represented what was typical for their audience. A nice even one-kilogram range from 3.0 to 4.0 kg would have included far more babies.

My children having weighed more than the seven-pound-eleven-ounce cutoff at birth, I was curious to see what risk numbers the authors’ analysis would assign to them. I found my answer in an online appendix to the study. For “cohabiting white women age 30+ with 1st baby birth weight > 3500g,” the expected rate of SIDS for room-sharing infants was 11 out of a million, and for bed-sharing infants 31 out of a million. Put another way, my husband and I were theoretically running an additional risk of about 1 in 50,000 by bed-sharing.

That’s about the risk of dying while playing soccer – something we also allowed our children to do.

And if I had it to do over again, the only change I would make would be to start bed-sharing earlier, before I reached that point of exhaustion where I was a menace to my baby.


 

 

The studies analyzed in this post are:

2013 meta-analysis: Carpenter R, McGarvey C, Mitchell EA et al. Bed sharing when parents do not smoke: is there a risk of SIDS? An individual level analysis of five major case-control studies. BMJ Open 2013;3:e002299.doi:10.1136/bmjopen-2012-002299

2009 German study: Vennemann MM, Bajanowski T, Brinkmann B et al. Sleep Environment Risk Factors for Sudden Infant Death Syndrome: The German Sudden Infant Death Syndrome Study. Pediatrics2009;123;1162.doi:10.1542/peds.2008-0505